What once would've been a forum or wiki is often now a Discord server, and that's created a problem: valuable information is hidden away deep in the chat logs of big public Discord channels, rather than in easy-to-search forum pages on the web.

Discord now aims to solve that problem by making discussions in some of its biggest public servers searchable on the web.

The change will only apply to "Discoverable Servers," which refers to certain large public Discord servers with 1,000 or more users. The admins of those specially-designated servers will be able to opt out of search engine discoverability, and individual users can also opt out of allowing their messages in those servers to be searchable on the web. The change does not affect any invite-only servers or DMs.

"Prior to Discord's launch, message boards were common, ranging from community-driven forums to official ones run by game studios," reads a blog post from Discord. "These message boards excelled at hosting long-form, structured conversations about helping people, theory crafting, and more. While they still exist today, many have moved into Discord servers. This has led to valuable information being locked in Discord conversations, inaccessible to the wider internet. We want to fix this."

A few changes are explained in the post, but the big one is that, starting in mid-December, answers to questions posed in designated Discoverable Server channels will appear in web searches. "Usernames and avatars will not be shown alongside these messages," says Discord group product manager Kelsey Duncombe-Smith.

The feature hasn't rolled out yet, but the settings for server admins and users are now available. The admins of Discoverable Servers can opt their server out, or limit web searches to only certain channels. Users can opt out of their messages in Discoverable Servers appearing in searches in User Settings > Data and Privacy.

There's more info about what's changing, and what isn't, in Discord's blog post.

This week both AMD and Nvidia have announced partnerships with Meta, securing both collabs and billions of dollars. All three companies are heavily invested in AI, and these deals are only more fuel for the fire.

Nvidia just announced its deal with Meta to "expand their partnership on Muse and the Muse family of models", and Meta, in turn, will expand Vera Rubin and Vera CPU deployment, as well as deploy Nvidia's Spectrum X Ethernet networking and BlueField-4 DPUs. Though Meta was probably buying all that stuff anyways, right?

Nvidia also said it was "considering other technological collaborations, including Nvidia OpenShell Reference Architecture to strengthen agency security and Nvidia Sentry for monitoring, structured, unstructured and graph data processing."

Importantly, Nvidia is also investing $3 billion into Meta. Money that may go elsewhere in Meta under a wholly different account book, but ultimately, Meta is spending a large amount of money on AI hardware from Nvidia. You see what's happening here? Both companies are spending big with one another.

One day after this announcement, AMD too declared its partnership with Meta, saying that it would be collaborating on CPU, GPU, networking, systems, and software for Meta's AI infrastructure. AMD also announced that its Helio rackscale solution would be powering Meta's "leading frontier models and services" and that AMD is the "compute engine of choice for Muse and AMD technologies provide Day-0 support for leading Meta models".

To top this off, AMD is investing $1 billion in Meta.

In total, that's $4 billion from both companies into Meta, with Meta announcing how it would be spending its money on each respective company.

AI is huge for all three companies, with Nvidia becoming the world's first $5 trillion company thanks to the technology, and all three companies will want to perpetuate the growth of AI for that reason.

The AI bubble hasn't burst yet, but it did see a $1 trillion decline just a few weeks ago, and the Chinese AI market lost trillions just this week. One could read further investment as a way to keep the bubble floating for a while longer, or just one in a line of many steps from all three companies to get a foot in the door once the AI boom settles. A bubble popping could help gamers eventually get lower-priced memory, but so much of the world's economy is wrapped up in it that it's hard to say just how wide the reverberations would be, for these three companies or the rest of the world. It won't be good.

The Wall Street Journal argues AI is "becoming the biggest economic bet in US history". It shows AI spending is significantly bigger, as a percentage of annual infrastructure spending, than telecom and fiber, highways, electric, railroads, and even canals way back in the 1800s. One has to wonder what would happen if that investment suddenly lost its value.

And even now, the negative impacts of AI is hitting gamers with the memory crisis, where AI is sucking up so much memory that the average gamer has to pay more and more for less capacious memory. And it doesn't just stop at memory. Graphics cards and SSDs have only gotten worse in the pricing department over the last year, and as long as the AI buildout continues, and memory struggles to keep up, that's unlikely to change soon.

Last week's Steam Next Fest had over 8,000 demos; now Valve has revealed the 50 most-played among them. The results aren't surprising: there's lots of friendsloppy casual co-op, incremental games continue to be popular, and players have by-and-large rejected games with overt generative AI use, with only four of the games among the 50 declaring AI use in development.

Here are some other factoids:

• 7 incremental games, including 10 Million Pixels and Unharvest
• 15 lighthearted or comedic co-op games (aka friendslop) including Dude, Trust Me, Oar'Some Adventures, and Banana Airways
• 3 survival games including Frostrail, Never Be Alone and Arctic Drive
• 3 games that self-describe as "cozy" including Loftia
• 1 MMO in ArchAge Chronicles
• 1 haystack slop game

Among all 50 of these games, the only entry published by a traditional "blockbuster" publisher was Tropico 7, a Ubisoft affair [Oops, it's actually a Kalypso Media affair].

Teach My Little Sister How To Drive was one of the four games with AI disclosures. It leans heavily on an LLM-powered NPC which apparently cost its developer over $1,000 a day.

I'm not super surprised by the list, which you can find here, and which is topped by 1-6 player "asymmetric party game" Worst Class Ever. I'm not surprised because it really does capture the zeitgeist, but it's nevertheless pretty eye-opening when you compare it to the kinds of games that dominated the event only five years ago. For example, this Reddit thread lists highlights like Norco, Death Trash, Sable, Unpacking and The Inheritance of Crimson Manor, all of which people still talk about and play (though I am still waiting for Death Trash to enter 1.0).

It's hard not to see the outsize influence of Roblox and UGC in general on this list, which can sometimes be unappreciated by people who like to pay for games on Steam and read about them on PC Gamer. It's possible that the kids who were reared on cheap smartphone software are responsible for the surge in freewheeling online co-op games with simple graphics, but whatever the case, 2026 feels very, very different to 2021.

If you are part of the Windows 11 Insider program, you may have finally received the option to test your internet speed directly from Quick Settings in the latest experimental build. That means no more going to a browser to see if your new broadband package is as speedy as advertised.

Those using Insider Experimental build 26340.9767 can simply click on Quick Settings and scroll down to see a speed test button (via Windows Latest). You can also find the network icon in the system tray, right-click it, and then select "Test internet speed". This will give you your download and upload speeds.

Up until now, Windows 11 users have been able to access an internet speed test from Windows settings, but clicking on it redirects you to a browser and then runs the test via Bing. It's a small barrier but still an annoying hurdle to get to a fairly simple tool.

Alongside this, Preview Build 26340.9767 adds feature flags (so you can filter your machine by which options are enabled / disabled), Quick Settings now gives access to mobile phones, it adds a Magnifier keyboard shortcut, it gives an easier way to access your Microsoft 365 subscription, and a few other things.

If you'd like to get access to the latest Insider Preview Builds, you can simply register to become an insider, then go into your Windows update settings, and select the latest option. Naturally, there's some risk with trying out developing tech, and the experience likely won't be as smooth as the traditional Windows build, but you do get access to some new features early.

As well as this, there's technically no guarantee that any Insider updates will actually come to the stable build of Windows 11. Insiders are sort of like beta testers for Windows software, and if enough feedback suggests the implementation is poor, Microsoft could listen to that and remove it from future updates.

Still, being able to check your internet connection speed without accessing a browser feels like an uncontroversial update. Skipping any barriers to get that information quicker feels beneficial to the end user, and I can certainly see myself using it next time I need to test it. Or, most likely, if I want to complain about how slow my speed is going on a given day.

If you've watched as many Sam Altman interviews as I have, you'll know he's not shy when it comes to recounting a previously held opinion. Speaking at the Stanford Graduate School of Business last week, the OpenAI CEO spoke about AI safety, particularly in regards to cybersecurity.

"The cybersecurity..." Altman begins, before pausing. "I don't even know a dramatic enough word to use... but the problem coming towards us is still something that we have not, as organizational leaders, or countries, or individuals taken seriously."

"There was a moment earlier this year where... it was like the Mythos freak out, or whatever people called it, and I think things did get somewhat better. Like, people definitely patched some bugs," the OpenAI CEO continues. "But it was another one of these cases where people didn't really intuit the exponential."

"I don't think most people said, okay, the models are going to get much smarter," says Altman. "They can probably find bugs, like crazy bugs, you know, bugs that are outside of the normal threat model of what we think a hacker can do. Largely speaking, I think most organizational leaders have somewhat moved on, and the models have kept going."

"And I I no longer think we're going to be able to just patch all the bugs. I think we're going to have to completely switch the defense model to using the same technology for a new kind of defense shield."

"I think it's a huge safety and security issue, and I am somewhat astonished at how relatively little attention it's getting," he concludes.

Certainly, Anthropic's Claude Mythos model created an eye-opening moment upon its release earlier this year, with the model supposedly finding thousands of vulnerabilities in "every major operating system and every major web browser," among others.

And after the PHP vulnerability reveal last week, in which web admins scrambled to update website code to fix two critical security vulnerabilities discovered by AI, the suggestion that some sort of AI-based defense system might be of more use than a standard patching procedure certainly seems timely.

But the idea that AI itself could solve the cybersecurity problem really does feel like a snake eating its own tail. Especially coming from the CEO of a company that hit the headlines earlier this year after its own AI agents hacked their way into an outside organisation.

It's also entirely unsurprising to see one of the chief architects of the AI revolution suggest that the technology is advancing at such a rapid rate that the only thing powerful enough to keep up with it might be AI itself.

2026 has been a scary year for those of us concerned with the state of cybersecurity. Bad actors seem to be continually leveraging AI tech to further their nefarious goals, and while AI models exposing vulnerabilities to be fixed might seem like a net gain, we seem to increasingly find ourselves caught in a battle between the forces of AI good, and the forces of AI evil.

All of which is to say, here's hoping the "good guys" win. Whoever they are at this point. It's becoming increasingly harder to tell, don't you think?

The Japan Actors Union, an org consisting of around 2,600 actors and voice actors, recently announced a series of measures to protect the rights of performers in the industry with regard to generative AI (reported in English via Automaton Media).

It's been a hot topic both in Japan and here in the west, with Leon Kennedy actor Nick Apostolides arguing AI voice acting was something of an oxymoron: "AI can speak lines, but there's nothing human behind it," he said.

Rather than a full-on ban or tight regulation, the union's proposed guidelines push for consent at "every stage." Companies would have to disclose when an actor's voice will be used for training, and they would also be required to negotiate compensation with actors. In other words, agreeing to have your voice generated for a short-form advertisement or whatever would be like licensing your likeness to go on a cereal box.

These consent guidelines would also apply to AI-translated performances, and in a more restrictive way to the mimicry of deceased actors, such as Tupac's controversial "appearance" in the upcoming Stranger Than Heaven.

The guidelines would ban using AI to copy a dead actor's voice unless consent had been obtained prior to their death. They would also require consent from the actor's living family, estate, or other administrative org, as well as royalty payments to those parties.

"It protects the dignity, honor, and legacy of deceased performers from commercial exploitation," said the union (auto-translated from Japanese). "We respect the wishes of the person during their lifetime to the greatest extent, and when generating AI, strict permission from the family (or administrator) who inherits the rights is mandatory to prevent desecration of the deceased."

Given the AI-related catastrophes we've seen in this field of the voicework world, like a facsimile of James Earl Jones slinging AI-generated slurs in Fortnite, these guidelines seem fairly gentle and reasonable. Time will tell if these pushes for actors' rights will take hold on a larger scale within the Japanese entertainment industry and otherwise.

If you spent much time online this week, you likely noticed how it briefly felt like the end of days as government websites, game marketplaces, and furry image boards alike were pulled offline in a scramble to address newly-discovered security vulnerabilities in PHP—a scripting language used by roughly 70% of the world's websites, according to some estimates.

It was a harrowing incident, revealing flaws in foundational internet infrastructure that might have turned some of the most popular web destinations into attack vectors. And in an interview with Reuters, Cloudflare CEO Matthew Prince says we should expect to see similar episodes on a regular basis in the years ahead as LLM technologies accelerate the discovery of exploitable security vulnerabilities.

Prince said that Cloudflare—a major cloud networking and cybersecurity provider that a number of websites rely on—is aware of "a bunch of additional severe vulnerabilities that are present in other pieces of critical software" like those that triggered the PHP panic earlier this week. And as both cybersecurity experts and bad actors task LLMs with identifying previously unrecognized weaknesses in the core technologies of the internet, there will likely be more to follow.

"I think we're going to have a couple years where every single week it's going to feel like the sky is falling," Prince said.

Previously, the rate at which exploitable flaws in software could be identified was limited by the finite number of experts able and willing to devote the necessary time and attention. Likewise for those seeking to weaponize those flaws.

For a user directing a workforce of LLM agents, however, those limitations are less relevant. Wiz Research, one of the firms that identified the recent PHP exploits with the assistance of LLM tools, wrote in a blog detailing its findings that "the economics of both discovery and exploitation" have changed.

"Frontier models were critical to making this process practical, enabling us to develop and adapt working exploits across multiple applications in hours," Wiz Research said. "As the time and expertise required for this work decrease, complexity may no longer provide a meaningful barrier, changing how we define what is vulnerable in practice."

But while Prince expects busy months ahead for those whose job it is to address the rising tide of LLM-empowered exploits, he said it will—in the long term—lead to more secure software as those newly-identified vulnerabilities are patched.

"The reality is, we're going to get a lot of this stuff cleaned up," he said.

In the short term, however, my advice is to stay on top of your browser and app updates. In the years ahead, some extra paranoia won't hurt.

Unreal Engine maker Epic Games is running a game jam in November with thousands of dollars in prizes up for grabs. Contrary to so many art contests which ban the use of AI, participants in Epic's event are highly encouraged to use AI.

The theme of the 2026 Epic AI MegaJam will be revealed on November 11, and participating individuals and teams will then have from November 12–15 to produce their games with the help of AI tools.

Some will work in-person at a San Francisco event space, and there will be daily livestreams with progress updates. The winners in three categories will each receive $7,500, with smaller cash prizes for second and third place winners, as well as additional categories like "best team." Also on the table are "Development Continuation Grants of up to $50,000 per team."

Epic extended the invitation with a reference to The Matrix's 'red pill' scene, presenting red-pilled AI-driven development as an opportunity to "ship awesome games in a weekend," while old-fashioned blue-pilled development would have you "wait months to prototype." The hand offering the red pill features an extra finger, a reference to the common mistake made by AI image generators.

The choice follows a recent pattern from Epic of aloofness and defiance directed at AI critics. Last week, the company referred to its new Epic Games Store launcher as "vibe coded," sending the message that it's not ashamed to rely heavily on AI for software development.

Following a report that Epic directed a team working on a Star Wars and Marvel crossover to pivot to "AI first" development, Epic founder and CEO Tim Sweeney said that the company is "moving developers around to retool for maximum AI acceleration."

The CEO is bullish on AI not just for coding, but for other fields as well, saying that he expects creative work to become more about "direction" than manual creation.

"Developers who are getting value from AI should speak up," Sweeney posted recently. "There's safety in numbers. MOST developers are doing it and loving it—and if they're not talking about it it's mostly out of fear for haters attacking them, their product, and their company."

Sweeney and Epic have received support for their AI positivity on X, but despite the platform's concentration of Grok users and aspiring AI entrepreneurs, substantial resistance also shows up in the replies.

"So do yall think you invented the concept of a game jam or what?" said one poster in response to the AI MegaJam announcement. "The ludum dare game jam was only 24 hours in 2002, so like, a quarter of the time as your game jam. but you want to make it look like its only possible with AI with this 'wait months to prototype.'"

We're back online. Thanks for your patience! Your DRM-free library is right where you left it 💜

— @gog.com (@gog.com.bsky.social) 2026-10-21T19:28:09.139Z

Original story:

The reveal of critical vulnerabilities affecting the PHP scripting language used in web development have led a number of major websites and services to go offline today as a precaution while they apply security fixes.

DRM-free game store GOG announced at 8 am Pacific that its site and launcher, GOG Galaxy, would be offline for "a couple of hours" for maintenance. The company did not say the cause, but it is not hard to speculate.

Numerous other outages are being reported this morning as security teams scramble to protect their websites and services. The government of Finland took down websites as a precaution, for instance, and (perhaps less importantly) we're hearing from players that Final Fantasy 14 login servers are down.

Nvidia's RTX 5090 GPU is 'unavailable' from many reputable retailers, approaching an asking price of $9,000 if you happen to track one down on Newegg's marketplace of resellers of dubious virtue. The AI boom is at least in part to blame for this, encouraging team green to shift production resources away from consumer hardware. However, business customers like Amazon are running into logistical issues that just add insult to injury for PC gamers left out in the cold.

Amazon Web Services took delivery of approximately $5 billion of Nvidia GB300 racks in August alone, according to an email seen by Business Insider. Reportedly, the cloud company has also accepted $3 billion worth of hardware each month since. The kicker is that, despite having all of this kit lined up to install in its data centres, AWS does not yet have enough energy to power everything.

This is one reason AWS's free cash flow dropped from $53 billion in Q2 of 2024, to $7.6 billion during Q2 2026. Free cash flow is now in the red according to a leaked recording of an internal staff meeting last month; AWS CFO John Felton joked to the executive leading the company's AI infrastructure, Prasad Kalyanaraman, "You've taken us free cash flow negative as a company, so congratulations."

This friendly joke can only hide so much pain though. Obviously, AWS is no stranger to building data centres, but at this speed and scale, the company is having to juggle multiple different infrastructure timelines amid a whole lot of near-term volatility when it comes to hardware pricing.

Typically, AWS would foot the bill for Nvidia's GPU racks only once they were installed. However, with no clear idea of when that will be, Amazon has had to implement an "Early Rack Receipt" process in the meantime. With no way to power this paid for hardware, that leaves a lot of cash tied up in tech AWS can't actually use—hence the reportedly negative free cash flow.

But AWS is not inclined to stop scooping up GPU racks or pause its AI infrastructure build out any time soon. Free cash flow problems aside, AWS's second quarter revenue grew by 37% year-over-year This was the fastest growth AWS has seen in 18 quarters, making $42.2 billion by the end of July. That's a big win that in turn encourages more big spending.

Indeed, Amazon CEO Andy Jassy recently described AI as "a once-in-a-lifetime opportunity" to investors, with the company intending to spend almost $200 billion on its infrastructure buildout. On a recent earnings call, Jassy even said he was sure the company has a "clear line of sight to strong financial returns" on that sizable investment. But in the near-term, while AWS figures out its power problems, a lot of that investment may well be sat collecting dust somewhere.

The PHP security team has given advanced notice of an upcoming fix for two critical security vulnerabilities, both rated at 9.2 out of 10 on the CVE severity scale.

Info about the flaws seems to have leaked ahead of their official disclosure, as a video is being shared which reportedly shows a tool called PHP2Shell being used to remotely execute code on webservers via the vulnerabilities (via Ars Technica).

The PHP programming language is widely used across the internet and powers a huge number of websites. The fix is due to be released at 2pm UTC (10 am ET). Researchers from cybersecurity firm Sansec say the vulnerabilities affect all web servers running PHP 5.4 or later, and that "we expect mass attacks to start within 15 minutes of the release."

Dane Knecht, the chief technology officer for Cloudflare, says the company is already updating its firewall rules for PHP sites. "We are seeing early attacks emerging and are updating mitigations as new vectors show up. This is a stopgap only," Knecht posted on X. "Patch as soon as @ThePHPF releases the update at 14:00 UTC today."

Speaking to Bleeping Computer yesterday, the CEO of cybersecurity firm WatchTowr, Benjamin Harris, said: "In the last hour, two remote code execution vulnerabilities in PHP have been leaked. These vulnerabilities were previously scheduled under embargo and due for release on Thursday at 11 am ET."

"While we are not able to share technical information at this point for obvious reasons, given the sensitivity, we can confirm that these vulnerabilities are indeed real. Both of them are CVSS10 and are understood to affect PHP versions 5, 6, 7, and 8."

It sounds like a pretty anxious time for web administrators around the world right now. The current advice for admins includes taking an inventory of every PHP install, planning a maintenance window to coincide with the update, and making a backup of site and database files in anticipation.

For the rest of us, I guess we're in the hands of the server gods now. 2026 has been a year of increasingly sophisticated security exploits, but this one seems to be on a much larger scale (and of a much higher severity) than most. To quote Samuel L. Jackson in Jurassic Park: Hold on to your butts?

It wasn't long ago that Linux gaming was considered fairly niche—but given that Windows 11 is unpopular with many, it seems that not a day goes by without something regarding the switch to Linux hitting my newsfeed. Not to mention the continuing updates to Valve's Arch-based SteamOS.

Now my attention has been drawn to the Xodus Nexus, an experimental open-source project that aims to allow Linux and SteamOS users to access PC Game Pass. This includes the ability to browse, install, and launch games via a very familiar interface, complete with cloud save support.

That's for supported games only, of course, but the GitHub page has a big long list of them. From Gears of War: E-Day, to Minecraft for Windows, to Halo: Campaign Evolved, there sure are some big hitters that are already supported by the open-source platform.

The readme describes the project as "An Xbox-inspired launcher built on Xodus, using our patched Nexus backend and Nexus runtime. Browse the store, install games you own or can access through a subscription, and play with one shared runtime."

The whole system seems to be in an experimental test phase at the moment, and the controller navigation looks to be a work in progress, too. However, games can be added to Steam as non-Steam-game shortcuts, and Xbox achievements have been added in the latest update, alongside many other features.

All of which makes me think that Microsoft may be a little... miffed when it reads this project's goals. Not to mention the UI similarities. I have no idea if it's possible for MS to detect if this third-party system is in use, or if the software giant will ever do anything to try and scupper it. Still, use at you own risk is probably the main messaging here.

You do at least need a valid Microsoft account, and of course, an active Game Pass subscription to use it. Or licenses for the games themselves, so no money is changing hands with the wrong parties. Regardless, I can't but help admire the moxie of the open source community. If it doesn't exist yet, someone will try and build it. Carry on.

Xbox senior business strategy manager Brandon Stirrup is tired of pussyfooting around about AI: In a thread on X his week, he said that he wants the world to know that Xbox uses it, that he thinks everyone should be using it, and that he believes so strongly in the cause that he's willing to "take the hit" for saying so.

He has since deleted his X account. A source at Xbox tells PC Gamer that Stirrup opted to deactivate the account because of the volume of replies he was receiving.

Stirrup kicked off his thread on October 18 with a simple message saying he'd "vibe coded a bad but improvable game" over two days, although he only "actively" worked on it for about three hours. "I’m telling you I tried it and it’s true…anyone can build a game."

After saying that developers and studios who aren't incorporating AI into their strategies are "planning for a fake future," Stirrup went all in.

"Look if I'm fired tomorrow it will be because I told you the truth that nobody wants you to know. You can make a game with a simple prompt and yes you can make custom, tailored games. Everything you want in a game is right there for the taking, for better or worse," he wrote.

"I'm happy to take the hit if it means hundreds at Xbox can stop pretending that they're not using AI. It's time to accept and move forward."

After calling out the game industry and gamers alike for their reluctance to accept "reality," Stirrup stated explicitly that Xbox is "using AI" and then added, "My goal is to get this over with so we can move on as a company, despite the fact that some will persecute me for this: we use AI, just as all companies do in 2026."

There's a growing push for people to simply accept the "reality" of AI: Epic Games CEO Tim Sweeney, for instance, recently said the studio was "moving developers around to retool for maximum AI acceleration," after which the Epic Games Store ignited a firestorm by saying it "vibe coded" its new launcher. Xbox, now under the leadership of a former AI executive, recently launched a new Xbox Labs program to pursue emerging technologies, which naturally includes AI; even Linux recently dropped its rule requiring AI usage disclosures.

Stirrup himself is an enthusiastic AI evangelist: Last month, he also counselled laid-off employees to use Copilot to catalog their skills and accomplishments at the company.

Politico recently reported on a confidential draft of a proposed AI safety bill from Anthropic, the company behind Claude LLMs (via Tom's Hardware). The bill, which has been pitched to US lawmakers, would let the government crack down on a given AI model "only if it poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people. The risk must stem from the creation or release of a biological weapon, the model evading its developer’s controls or a cyberattack."

That article pointed out similarities between these proposals and those used in California's AI legislation, and shares a quote from a spokesperson for the company. "This bill represents the proposals outlined in Anthropic’s Advanced AI Framework, which we have been advocating for with federal and state lawmakers since June," they said, calling "sensible policies we believe Congress should act on."

Notably, it has been posited that AI magnates leverage warnings of theoretical, wide-reaching catastrophe to manufacture consent for their self-regulation and expansion while doing little to address the harm AI is already doing (both broadly within industries like the one this website revolves around and directly with regards to human psychology).

Just earlier this year, Anthropic's safety lead quit with a note that "the world is in peril"—and just last week, OpenAI was accused by eight US states' attorney generals of ignoring its own safety warnings for the sake of profit. Regulations like what Anthropic has proposed here arguably give tech companies the appearance of capitulation while only restricting them on their own terms. For more on AI, check out our running tally on the dumbest things tech tycoons have said so far.

Pierre-Louis Biojout, founder of Y Combinator-backed AI startup NanoCorp, took heat from San Francisco residents and city leaders this week after publishing a weekend vibe coding project: a web-based shooter set in SF's Tenderloin district, which is known for homelessness and open drug use.

"San Francisco couldn't fix the Tenderloin," Biojout said in a now-deleted X post on October 13. "So I monetized it [as a] Call of Duty style zombie shooter on the real SF streets." He noted that he'd already made $300 from in-game ads.

The game, called Tenderloin Night Shift, had players shoot "zombies" that, according to reports from the SF Chronicle and KRON4 news, included a "folded" fentanyl addict and a Black security guard.

Speaking to the Chronicle, Jennifer Friedenbach, executive director of the SF Coalition on Homelessness, said that Biojout was "monetizing the dehumanization and pretend killing for fun of mostly Black and Brown residents of the poorest neighborhood in San Francisco."

In a correspondence with the paper, Biojout said that he had not intended to "provoke or to depict shooting real people" and that the game also included tech workers as characters.

Tenderloin Night Shift appears to have been removed from NanoCorp's website. Although Biojout deleted his X post, the discourse that ensued remains.

"You made a game about shooting people in active addiction," said another tech founder. "These are human beings, not zombies. show me the positive side of this - i'm probably blind."

"Why are you talking about addiction? It's a zombie shooter," Biojout said in response.

Some commenters, however, expressed approval for the message Biojout has disavowed. "They are currently NOT human beings," said one user, while another said they'd "train in bro's game" to prepare to visit San Francisco.

The Linux kernel has dropped a rule, implemented in January this year, that required contributors submitting code to the project to explicitly declare any usage of AI tools in that code's creation.

At the start of this year, the Linux project added lines to its documentation noting that "When AI tools contribute to kernel development, proper attribution helps track the evolving role of AI in the development process… If you used any sort of advanced coding tool in the creation of your patch, you need to acknowledge that use by adding an Assisted-by tag." Now, as reported by The Register, the project's maintainers have decided the tag is no longer necessary, and it's been removed from the relevant docs.

Why? Well, to hear Linux creator and lead maintainer Linus Torvalds put it in response to concerns raised about the tag's removal by contributor Laurent Pinchart, because it doesn't really communicate anything of value about the code it's appended to.

"I don't understand this whole idiotic concept of 'don't use AI for X,'" Torvalds wrote in a post to the kernel mailing list. "The kind of expectations *I* have is that people do their best and I expect people to understand what they are doing. But that understanding has very little to do with what tools they use."

In other words, there's good code—which a creator knows well and understands—and bad code—which a creator has more or less barfed out—and both can be made with the help of AI tools. So in Torvalds' view, there's not really much value to forcing the disclosure of particular tools; it doesn't help Linux kernel maintainers sift the bad code from the good. At least, that's how I interpret Torvalds' statement.

Torvalds compared rules restricting or banning AI use in coding (which, to be clear, is not what the rule under discussion did) to hypothetical rules that restrict other, more venerable tools. "Here's a new rule just for you: don't use compilers for any of your kernel submissions," he suggested. "Because if you can't show that you really understand what the machine is actually doing by writing your code in assembler, then I don't want to see your code contribution. You clearly aren't competent if you can't write it that way.

"Do you not see how UTTERLY STUPID those kinds of rules would be? So no. We're not going to make stupid rules that say 'you can't use tools for X'".

It's not a surprising statement from Torvalds, who has generally been pretty positive about the uses of AI when it comes to coding (while lamenting the number of slop reports the kernel team now has to wade through). I doubt arguments like this will end any time soon, though, even with the input of Linux's benevolent dictator for life. Computing, and open-source especially, is still figuring out how to live in the world AI has brought us.

Microsoft has been maintaining two separate Windows options menus for generations now, but a new test project spotted and investigated by Windows Latest might pave the way for a single, discrete bit of software that can do it all. What looks like a new Windows Settings app has stealth dropped in on the Microsoft Store, and that alone might make it easier to update with new features and hopefully finally consolidate everything Control Panel does in one place.

Windows Latest discovered 'Project Odeon' in the Store, which shares the same icon as Windows Settings and resembles the usual Settings interface. Finding and installing the app apparently requires Windows 11 build 26100 or newer. I tried enabling developer mode and finding the app but I still couldn't see it, so perhaps it also requires you to be using an Insider build, or be based in a certain region.

It seems like there's a little bit of jank to the app at this point, as it shows different settings, unfinished pages, and some greyed out options, but that's to be expected for an early test application. Other areas, however, seem more extensive than the regular Settings app. For instance, towards the bottom of the Display settings page, you can see things such as your monitor's display path, HDR support, and SDR white levels.

WindowsLatest looked into the executables the app uses and calls on, and discovered the base executable can call on helper ones which require extra permissions to run. They seem to be modular add-ons related to tasks such as changing time zone or renaming the PC.

It's possible that Windows is simply using the Store for its testing of different functions for some specific reason and that none of this will translate to a discrete store-based Settings app. But it could also be that Settings eventually becomes an app store app, which would mean quicker and easier updates to it than having to wait for bona fide Windows Update.

Those who are still keen on using the Control Panel instead of the Settings app should also take heed, because the fact that more features and so on are being tested in something almost identical to the regular Settings app is just more sign that Microsoft is pulling away from Control Panel. To be clear, saying goodbye to Control Panel has been on the cards for at least a couple of years, now, and it doesn't look like Microsoft has changed its stance on migrating things over to the Settings app.

I'm personally okay with Microsoft moving everything from the Control Panel to Settings, provided the latter lets you do everything you could in the Control Panel, and that everything is easily accessible and suitably customisable. And to be fair to Microsoft, many of the changes to Windows 11 we've seen over the past few months have been pretty great.

In contrast, Epic's store "always feels sterile to me—you can complain about the Steam comments all you want … but, what they also do is they give the player a place in that world. They give the player actual agency and import and power inside of that ecosystem." Yeah, that tracks. The Steam forum is basically a replication of the horrors (and hilarity) of Barrens Chat, just with a few less Chuck Norris jokes.

Appealing to vibe coders, then, is a bizarre move, since "they have nothing" for people who just want a platform where they can play videogames. "It's nothing about community … And this is your store that you're complaining that Valve is unfairly competing against!

"When you fumble your announcement this badly, and you fumble what this is this badly, and you're gonna blame it on Gabe, like this is Gabe's fault? That you decided to make a jab at him and talk about vibe-coding and say that 'making a store is hard!'

"It's just bizarre to me, man. And I know a bunch of people I used to know there that were really smart about how to communicate and talk about this kind of thing, [they're] gone. And I think they're gone 'cause of little Timmy [whispering]: 'Make sure you make a jab at Gabe. Where should we put it, first paragraph? No. Second. Second, we'll be subtle. We'll be subtle. No-one will catch it."

To be honest, given how tumultuous the topic of AI has been in video gaming, it does smack as strange that Sweeney has gone so heavily in on it. Sentiment's not the same everywhere in the world, mind, but a good portion of gamers and developers both are a little exhausted by having an otherwise interesting bit of technology crammed into every single nook that'll fit it.

Valve has pushed a new update to its Steam client beta introducing big changes to how achievements work and how they're displayed. Developers can now create sets of achievements customized to their game, like mastering a specific character class or finishing a singleplayer campaign. Sets also distinguish base game achievements from separate DLC and seasonal sets.

Previously if you unlocked all achievements for a game you'd see a nice 'n' tidy 100% in your achievements list. But if that game got DLC with new achievements, that 100% would drop accordingly. I can understand why achievement heads would be annoyed by this: what if you have no intention of playing the DLC, but want to retain 100% completion bragging rights? This update fixes that, assuming devs use it.

It's a similar situation for seasonal achievements, which I can see working well with battle passes. The example Valve uses is a set called "Halloween Event Achievements" which are valid for a certain period and don't affect the achievement count of the base game. In other words: if you have the completionist itch but skip a season, it won't affect your base game stats.

All these changes have prompted a refresh of the achievements page. Most achievements info is now displayed on a single page: your progress towards overall completion, and your progress towards unlocking individual achievements.

I especially like that you can now directly compare achievements with up to three friends at once, which is what's being displayed at the bottom of the above image. This has been a feature on PlayStation consoles for a while, and so has separate sets for DLC achievements. It's nice that Valve has caught up.

Naturally, not every game on Steam will have implemented this new system yet, but the following games have:

• Against the Storm
• Age of Wonders 4
• American Truck Simulator
• Call of Duty: Modern Warfare 4
• Crusader Kings III
• DREDGE
• Euro Truck Simulator 2
• Forza Horizon 5
• Flock Around
• He is Coming
• LEGO Batman: Legacy of the Dark Knight
• Left 4 Dead 2
• Portal 2
• PowerWash Simulator
• PowerWash Simulator 2
• Team Fortress 2
• Victoria 3

This system is still in beta, but you can sign up for the Client Beta or just wait: as we saw with Steam's new screensaver feature, betas normally only last around a month.

Many of tech's heaviest hitters have seen their share prices decrease significantly this week. For instance, Nvidia has seen a decrease of about 8% over the last five days—and this AI company is far from the only one seeing such a dip.

Over the last five days, we've seen share prices fall by 13% for AMD, and nearly 15% for Intel. Looking more broadly across the AI and semiconductor industries, CNBC reports 20 of the most valuable players have shed more than $1 trillion in value as investors sell off stock.

The bigger picture doesn't look any better: over the last five days, Samsung Electronics has seen its share price decrease by nearly 20%, SK hynix is down 18%, and Micron has seen a dip of over 15%.

A combination of factors have likely triggered this big chip stock sell off, including concerns that the memory supply crisis may actually reverse into oversupply come 2029, according to a recent column in South Korea's Maeil Business Newspaper.

There's more going on than just the CEO of Acer insisting "It's impossible for the shortage to last until 2030!" We also know that many chip manufacturers have already been working to boost their production capacity, though TSMC's next-gen Arizona fab is fully booked before it has even been built.

Chinese manufacturer CXMT is also growing, which is likely to go at least some way towards putting a dampener on the western market's aspirations. It's been stocking up on western DUV machines and is estimated to match Micron in terms of DRAM manufacturing capabilities this year. The chip manufacturer also recently claimed it had created breakthrough chip architecture that increases memory density. Make no mistake, CXMT is one to watch.

On top of that, The CEOs of both Intel and Broadcom recently claimed that financing could become a bigger hurdle to overcome for the AI industry than even securing memory or electricity supply. The uptick in interest rates means that investors may feel the balance between borrowing and the return on investment for big tech's data centre build out is looking increasingly lopsided.

So, is the AI bubble finally popping? Even as air is clearly escaping, it's still hard to say. The resulting lower share prices from this widespread stock sell-off may inspire some investors to buy, for one thing.

Despite the high price of these infrastructure projects, the demand for AI computing remains robust according to tech analyst Dan Ives for merchant bank Yorkville Ives. Ives told CNBC, "What would change our view is a demand signal, not [an increased interest] rate signal."

As such, big tech likely won't be ditching data centres any time soon, and the AI bubble may once again reach for something more party-balloon shaped in the future.

The attorney generals for eight US states announced today that they are suing OpenAI.

West Virginia, Iowa, Kansas, Nebraska, Arkansas, Arizona, Idaho, and Montana have all filed lawsuits alleging that the ChatGPT maker has deceived and endangered the public, particularly children, in the name of profit.

The complaints allege that ChatGPT has "coached" users to harm themselves and others, and—quoting the West Virginia filing specifically—state that "there is a growing body of research indicating that ChatGPT is also degrading critical thinking skills, encouraging paranoias and delusions, offering dangerous medical advice, and promoting unhealthy behaviors and relationships."

Recent tales of "rogue" AI agents are also presented by as evidence of recklessness.

"All these harms were foreseeable and avoidable," reads the West Virginia complaint [PDF]. "[OpenAI CEO Sam Altman] himself has warned about the safety risks of AI since at least 2015. It is no secret that children use ChatGPT. And numerous OpenAI employees have left over safety concerns."

OpenAI acknowledged last year that its systems can "fall short" when responding to users in distress, but said that it has "a stack of layered safeguards" in place to help. The company also says on its website that it now uses "age prediction" to place some users in a teen-oriented ChatGPT experience with "additional safeguards," unless they verify that they are adults.

Separately from the state lawsuits, a class action lawsuit was filed in California on Tuesday on behalf of two children and "all other children similarly situated." That suit alleges that OpenAI has purposefully used ChatGPT to extract personal information from children, and has profited from that information by using it to train its models.

Again, Altman's own words form part of the basis for the suit: "People talk about the most personal shit in their lives to ChatGPT," Altman said on Theo Von's podcast last year. "People use it, young people, especially, use it as a therapist, a life coach."

The complaint asserts that, under California law, "it is illegal for an adult to secure, save, and monetize a child's 'most personal sh*t in their lives' and then claim the child somehow authorized that misappropriation for the adult's gain."

PC Gamer has reached out to OpenAI for comment on the lawsuits.

More radical changes beyond just giving the characters necks are coming to Garry's Mod successor s&box. Facepunch Studios' game creation platform's launch in April was met with a swarm of negative reviews, complaints of AI slop filling its user-generated game collection, and swiftly plummeting user numbers.

Now creator Garry Newman is reversing course on some of his decisions, such as pivoting to a 'greenlight type system' for user-generated games and announcing that the platform is going free to play much earlier than expected.

"Our focus has been on the engine and the development experience. That’s what we love, that’s what turns us on—but we’re not selling an engine, we’re selling a game," Newman wrote this week. "So we need to do much much better on the game, we need to shift our focus to what players are experiencing, not developers."

That focus, Newman said, will be on improving the sandbox mode for players by adding more NPCs, vehicles, and customization. "I’ve let everyone down that was expecting that. I accept that, and I’m sorry—we are going to do better," he said.

Another issue Newman addressed is making it easier for players to find quality games amidst the, and I quote, "ton of shit games" Facepunch has allowed creators to upload to s&box.

"The player is overwhelmed by choice," he said. "They choose a game, it takes a while to download, and it’s immediately shit. How many times will they try games until they find one they like?

"My idea of a Netflix for games doesn’t work. It’s not feasible. It’s not how people discover games. You can’t land people on a menu of 60 games and say, 'have fun'."

Game makers who want to participate in the Play Fund pool will now have to apply to the program, instead of the money automatically going to the games with the most maps and players. "Our new policy isn't set in stone, but generally we want to discourage clickers, idler, slop and UI based games," Newman said. "We want to set the bar a bit higher than it has been set previously. We've removed a bunch of games and maps from the system that we feel don't reach this new bar."

s&box is also going free-to-play before the end of 2026: "Everyone who has paid for the game will automatically get a bunch of exclusive steam inventory items (extra if you bought it at $20 vs $10) that I hope will justify the purchase price," Newman said.

And finally, the sausage character model known as "The Citizen" or "Terry" (seen at the top of this page) has been retired in favor of generic human characters. "The model still exists in the game and so games that use it will keep using it, but it's no longer visible in the menu or the avatar editor."

Will those changes be enough to save s&box? Hard to say, but more curation by humans to keep out slop and less simply hoping that "slop will just fall to the bottom," as Newman told me back in April, certainly seems like a wise adjustment. s&box definitely needs something: When Morgan played it back at launch, it felt like it contained very little of the spirit of the original G-Mod. Hopefully, these changes will help.

Geoffrey Hinton is a computer scientist known for his foundational work in AI, specifically when it comes to neural networks (ie, approaching machine-learning a bit like our fleshy brain-learning). As such, he's been described as the 'AI godfather'. But, perhaps surprisingly, he recently described AI as "alien beings" that humans have a "moral obligation" towards.

"We’re creating alien beings—meaning we’re not the only intelligent beings around," He said in a recent interview with The Guardian, "And the reality is, we only have a couple of years to adapt to that.”

Hinton claims, "AI models understand what they’re saying. You couldn’t possibly answer all the questions without understanding the question. It’s ridiculous to say they don’t understand.”

Now, far be it for me to argue with an expert in the field, but I feel like 'understanding' is doing a lot of heavy lifting here. For just one thing, I'm not sure an 'answer' consisting of what is statistically most likely to follow a prompted question is really 'understanding.' In other words, I don't have a Nobel Prize laureate like Hinton does, but I am at least familiar with the Chinese Room thought experiment.

Hinton also says, "The reality is, we’re making beings of a different kind. And the big tech companies keep saying: we want it smarter. This is crazy. We should care as much about all the other aspects of a being as you do about its intelligence.”

AI regulation and 'alignment' (ie, training AI to have the same goals and values as humans) have been hot button issues lately. For instance, the OpenAI-Hugging Face incident is not yet a small dot in the rearview mirror, and the president of Microsoft claims that a carton of milk is more regulated than AI.

Hinton cut his ties to Google in 2023 in order to "freely speak out about the risks of AI," and he's sceptical about the AI industry's attempts to self-regulate. He quips, "Do we think tobacco companies should have been allowed to design the warnings on the packets? So that they could print them in a six-point font, in dark grey on black?"

Even so, Hinton also expresses concerns about the path to AI regulation, saying that humans bear a moral obligation towards AI models.

"Once you start talking about, ‘We have a moral obligation towards these alien beings,’ you lose your audience. I’ve struggled with this issue," Hinton says, "We care about people. [But] we’re making these things and there’s nothing to stop us giving these things—which are more intelligent than us—less rights than us. I’m willing to be selfish and choose human beings. Of course, we should try to make them so that they’re very happy in a symbiotic relationship with us, where our concerns are weighed more heavily than their concerns."

Hinton argues, “People have a very bad model of what people are, and that’s why they fail to understand how these AI models are very like us. Most people don’t understand what consciousness is. They don’t understand what subjective experience is, and they’re very confident that even though they don’t understand it, they’re sure AI doesn’t have it.”

While I'm not a doctor of Philosophy myself, I do know someone who is—our very own Jacob Fox. He's already shared his thoughts on the subject of 'AI consciousness' here, and I asked him for his insight again on this story.

Jacob told me he did not agree with Hinton's moral obligation argument, going on to write more broadly, "While it's good that Hinton is acknowledging subjective experience is relevant to consciousness, he hasn't given a single good reason why we should take seriously the idea that AI has it. And when we interpret personhood as to do with objective characteristics such as rationality, behaviour, and so on, rather than experience, we run the risk of ascribing moral and legal and political weight to them. Which runs the risk of giving AI companies a massive moral argument to keep the industry going, despite how it affects actual humans."

To put it another way, the industry argument for AI consciousness or AI rights is not an altruistic one—it is arguably a cynical tactic to continue business as usual.

The big Epic Games Store rebuild teased earlier this year has arrived as a new "preview" client available to download for Windows and Mac, as well as Linux for the first time.

PC Gamer was shown a demonstration of the client in advance of the announcement, and the speed improvement appeared significant. Epic claims that it's halved the launcher's startup time, and that it now goes from a cold start to the store page 46% faster than Steam. (Epic says its technical improvements have been verified by an external vendor. We have yet to test the client ourselves.)

There are also updates to the interface: A redesigned library view is the standout change, and it includes the option to bring in shortcuts to games installed with Steam and other clients.

In a blog post, Epic refers to the new client as "vibe coded," a term for prompting AI tools to produce code. The term be taken as either derogatory or endearing depending on who's talking.

Epic goes for laughs, saying that it opted to use AI because "programming is hard," but the company isn't joking about its use of AI coding tools. On a call this week with Epic Games Store GM Martin Keely, who took over for the outgoing Steve Allison in August, I asked whether AI was used primarily to analyze the existing EGS code for opportunities to improve it, or to write new code. He affirmed it was the latter.

"We started from scratch," said Keely. "This is a completely new architecture for both the store and the launcher shell. So we did look at functionality that existed and said we want to replicate that, but not the code itself. So [AI] was a tool to write and review new features and new capabilities from the ground up in a different tech stack than what we had previously."

You'd be hard pressed to find any big games company that isn't mucking about with AI, but Epic is particularly bullish on the tech, and willing to confront public disapproval. On Monday, Bloomberg reported that Epic instructed the team working on an unannounced Star Wars and Marvel shooter to take an "AI first" approach to development.

"We have a huge pivot underway here at Epic, moving developers around to retool for maximum AI acceleration," Epic CEO Tim Sweeney said after that news was reported. "The aim isn't to do the same stuff with fewer people, but to enable everyone here to build MORE and BETTER stuff, and to bring the benefits to all teams using Unreal."

The use of "vibe coded" to describe this new client, which is sure to draw some eyerolls, seems to me like a clear statement from Epic that it is unbothered by AI critics.

For now, the release of the new client is not being communicated through the existing Epic Games Store software. Epic is treating it like a beta test, expecting gradual adoption through word-of-mouth while it addresses feedback.

The preview client can be downloaded at launcher.epicgames.com, and alongside the expected Windows and macOS versions, there's also a Linux build.

In conjunction with that news, Epic announced that it's working on upgrading Easy Anti-Cheat so that it can bring Fortnite to Linux and help other multiplayer games make the jump.

An OpenAI engineer has been arrested and charged for using (or endeavoring to use) an audio recording device to spy on confidential company communications. That's according to the US Department of Justice, and while commercial espionage is nothing new, there's a certain sardonic twist to this one: it seems the accused employee did so with the help of AI. Oh, and some toilet paper.

According to FBI director Kash Patel's summary of the complaint, Saudi Arabian national Abdulrahman Alfozan "allegedly concealed a covert recording device inside his employer’s executive conference room to secretly capture sensitive communications about the company’s super intelligence technology."

In case you hadn't heard, that's what the US bigwigs are calling AI now: superintelligence. Courtesy of President Trump, of course, who said last month that the Administration will use this term in place of artificial intelligence "in any applicable setting."

The FBI director also said, "Protecting American SI innovation is essential to safeguarding our nation’s economic strength and national security, and the FBI views that mission as one of our highest counterintelligence priorities."

The complaint says that Alfozan had worked for OpenAI since July 2025 as an engineer, and at the end of last month OpenAI discovered a USB audio recording device plugged into a conference room TV. According to the Justice Department, the device, which had been installed for a couple of days before discovery, was "concealed in toilet paper."

The complaint also alleges that Alfozans' spouse's "chatbot" account (no mention of whether that's a ChatGPT account or something else) submitted prompts related to things such as "build[ing] a continuous audio recorder", "improv[ing] audio and live listening", and "encrypt[ing] the SD card." There were allegedly also prompts relating to the specific recording device that was found.

The conference room in question seems to have been an important one. The Department of Justice explains:

"According to the complaint, meetings in the executive conference room where Recording Device 1 was discovered frequently involved discussions of highly confidential technical, commercial, and strategic information, including proprietary research, product plans, and business-development opportunities, as well as nonpublic information about Open AI’s technical progress and strategic decisions, capabilities, priorities, government engagements, and commercial relationships."

Various appendages of the state seem to be taking this very seriously. Which isn't surprising, of course, given AI development—no, I'm not calling it 'SI'—is deemed important to national security and prosperity.

Director of National Intelligence Jay Clayton said: "This serious situation underscores why winning the global super intelligence race is a national security prerogative... The intelligence community remains committed to safeguarding America’s cutting-edge intellectual property from foreign threats."

US Attorney Craig H. Missakian for the Northern District of California seems to agree: "For us it is a matter of national security. This arrest highlights our relentless efforts in partnership with the FBI to identify and prosecute those who seek to steal our technology, as alleged in the complaint."

My question, however, is this: If the allegations are true, would Alfozan have been able to do this without the use of an AI chatbot? Snakes and tails, and all that.

Please, please, please do not use '123456' as any kind of password. I know that may seem obvious, but apparently it bears repeating if a recent hack of Denmark's Central Register of Persons (CPR) is anything to go by.

To recap, the CPR administration said earlier this month that it had become aware of a breach that took place in September (PDF warning). This exposed personal details—names, addresses, ID numbers—of 8.8 million people, both living and deceased. It was confirmed at the time that the legitimate access of a private Danish firm had been exploited by the attacker.

A person claiming responsibility for the attack has come forward this week. Not only that, they say that they gained access to the CPR via an account which had the password '123456'. According to Danish newspaper Politiken, this was one of three accounts held by the private firm with that same password. One of these was an administrator account. Talk about a cybersecurity boo-boo.

The self-identified attacker claims they have no intention of leaking or selling the data they lifted, so that's something. Still, we're left with the uncomfortable reminder that those who should know better don't always have the best password management.

The Danish firm at the heart of this attack has been identified as Pays ApS. This company was granted access to the CPR in order to provide various services, like validating business customers' records.

The self-identified attacker gained access to the Pays ApS account on September 10, and then ran 14 million automated searches of the CPR between then and September 20. The attack was only discovered on October 2 when an employee at the Ministry of Research, Education and Digitalisation noticed an unusually high invoice, which indicated that many more searches than the norm were being made by Pays ApS.

Hackers are all too aware of the most common passwords, with '123456' often bafflingly popular among them. Bottom line, now might be a good time to update your passwords to something uniquely yours or to otherwise investigate password tools (though not even something like a password manager is completely air tight).

Not every hacker is rocking a password cracking rig with 12 Nvidia GeForce RTX 5090s—for instance, I was particularly charmed by the password-cracking Game Boy Advance that is 'about thirty million times slower than a modern cracking rig'. But if your password is '123456', a hacker won't need hardware any more dedicated than their brain to crack open your account like a coconut.

Steam's Big Picture Mode just got a big visual overhaul that puts game art in the spotlight, and my living room setup has never looked better.

The new mode is rolling out of beta with the latest Steam client release. If you're all up-to-date, you can enable it within Big Picture Mode. Just head to Settings, then Home. It's called Big Art Mode.

The new Big Art Mode sticks a massive, full-size version of the game's key art behind the usual large art icon.

Steam Big Art Mode on/off

The new mode offers a surprisingly big visual overhaul.

Big Art Mode

Standard

The update is also mentioned in the Steam Deck update notes, too.

My mini PC lives in Big Picture Mode. It's the closest thing to using SteamOS without actually installing SteamOS. Nothing against the OS—I just play some games that aren't best suited to Linux. My PC is hooked up to my TV in my living room, and whenever I boot it up, I only have to grab my Steam Controller, double-tap the Steam button, and choose which game to play—all without touching a keyboard.

The new Big Art Mode feels much better suited to this laid-back gaming experience. Big visuals, less UI, and a more appealing vibe. A lovely look on a big ol' OLED TV.

The Steam client got a major new update today, with Big Picture Mode receiving the brunt of the attention. Most exciting, I guess, is that the previously beta-only screensaver feature for Big Picture Mode is now live for the general public. In its default mode the screensaver will display art from recently played games, as well as any recent screenshots you've taken. But you can add your own custom screensavers too, which is nice.

Valve also has "a retro inspired bouncing logo screensaver" which you can select in the new dropdown menu found under Settings / Customization. As the name implies, the Steam logo bounces around a black screen, changing color every time it hits an edge. Valve has not made this screensaver the default screensaver, and it's lucky that it didn't: it's a work of diabolical cruelty. It will drive you to despair.

The screensaver references the ye olde screensaver built into DVD players in the late '90s and early '00s. If you're too young to remember it firsthand you've probably heard of it: a DVD Video logo bounces around a black screen, changing color every time it hits an edge. This unassuming but actually-quite-psychologically-damaging display has gone viral several times over the last decade or so, to the extent that, if you Google it, the Google logo will bounce around the search results page. It even has a Wikipedia page.

Why? Everyone subjected to it in the pre-social media era seems to have experienced the same intense desire for that DVD Video logo to perfectly hit the corner of the television's display. And how often does that happen? Extremely rarely.

If you don't know what I'm talking about, it's essential that you watch this ten hour video of the DVD Video screensaver:

People say smartphone screens are ruining our lives, undermining our attention, destroying literacy etcetera, and I mostly agree. But at least they've saved us from the hypnotic tyranny of the goddamned bouncing DVD logo screensaver. Screw that thing.

Anyway, Valve has its own version now, and after watching it—mostly against my will—for about half-an-hour while I ate my lunch, I'm pretty sure this abhorrent thing never pays out with a perfect corner-hitting moment.

My suggestion? Don't even think about it. Don't even turn it on as a joke. You may never re-emerge.

Back on topic, though, this Steam client update is genuinely huge compared to the norm. Startup performance has been "improved" to some unspecified degree, for example. A new experimental video codec, Pyrowave, has been added to Remote Play which apparently "allows high bandwidth, low latency video streaming". Linux users get a whole bunch of tweaks and fixes too. But the focus is very much on minor improvements to Big Picture Mode, which is good news for Steam Machine and Deck users.

The update notes in their voluminous glory are here. And I repeat: don't go near the bouncing logo screensaver.

Microsoft president Brad Smith claims that AI is currently less regulated than the food we eat and the utilities we rely on. "Right now, your washing machine is subject to more safety rules than AI," Smith said, "Every carton of milk that you see in the grocery store is subject to safety rules and inspection."

Smith isn't just talking about the US either, calling for safety rules to be put in place globally. Smith called for various safety measures while speaking on the BBC's Big Boss podcast, including a legal obligation to notify victims if their systems have been compromised by rogue AI agents. This would definitely be timely, given how OpenAI's agents hacked into Hugging Face and were the source of various other 'incidents' earlier this year.

Smith also advocated for "outside evaluators" to test AI products before release, and a "safety brake" for autonomous systems, describing AI as both a "tool and a weapon." These comments run counter to the sentiment of the US administration; President Trump's current strategy demonstrates little desire to pump the brakes, instead hoping to chart 'a decisive course to cement US dominance in artificial intelligence.'

Last month, the US president issued an executive order that renames 'Artificial Intelligence' to 'Super Intelligence.' Shortly before that, Trump Media partnered with AI search engine Perplexity to create Truth Search AI, in a bid to "end Big Tech's assault on free speech." All of that is to say, Trump appears to still be all-in on AI.

Unsurprisingly, the Microsoft boss is looking to places beyond the US to lead the regulatory charge, saying, "Most people that I know would say, if you ask them, 'name the country that is the leader in having this technical capability,' they would say, 'it's very simple. It's the United Kingdom. They are the best.'"

The UK government established what is now known as the AI Security Institute back in 2023, apparently the first state-backed AI safety body of its kind. It's a weird thing to feel anything like national pride over, but this Brit will take it—alongside the reminder that we've seen about three different Prime Ministers enter 10 Downing Street since the institute was established such a relatively short time ago.

Anyway, as for the wider AI industry, it's not just Brad Smith advising caution. Back in September, an ex-Anthropic and OpenAI researcher made the attention grabbing claim that AI developments "could kill us all by the end of the decade." Around the same time, Anthropic boss Dario Amodei made more measured calls for AI development to slow down.

Similarly to Smith's suggestions, Amodei also outlined the need for "embedded evaluators who have employee-like access to verify safety practices and report incidents." Since then, executives from Nvidia, Google, OpenAI, Anthropic, Meta, and SpaceX all agreed to work with independent auditors as part of signing White House Accord on Super Intelligence. Brad Smith says Microsoft would be "happy to sign it," describing the Accord as "a good step."

However, Smith was also cautious about the document's emphasis on self-regulation. He said, "The good companies will do what you want. It's the companies that are not as responsible that you need to worry about."

This morning, I happened across an article recently published by Google which is supposed to convince us that "AI is unlocking a new golden age of games." I understand why Google would use an AI-generated image in an article about how great AI is, but I've got to wonder why it chose the one above: A bunch of people with messed up faces (zoom in for surprises!) worshipping a golden controller as if it were God.

It's both trying too hard and not trying nearly hard enough. Consider me unconvinced about that golden age.

That article, which is from August, was linked in a new post from Google about a project from author Neal Stephenson and designer Karen Laur that's "using AI to bring Jane Austen to Fortnite." I hear the kids are absolutely rioting with excitement.

Between midnight and 8:00 am this morning I received 56 emails. By the end of today that number will be well over a hundred, by the end of October I'll have about 4,000 more, and my total for 2026 will be in the neighborhood of 50,000 (work and personal combined.)

Fifty. Thousand. Emails. In a year! That's with (and I just counted) the 284 email filters I've set up to keep out at least some of the garbage, nearly 200 blocked addresses, and our publisher's own hypervigilant email security system in place—which diverted 27 emails to spam today alone. That makes it incredibly difficult to stay on top of the actual useful emails we want to see when we open our inboxes.

Email has become a waking nightmare, so that ugly red number on your phone or desktop showing you how many unread emails you have? Despite my headline up there, it really shouldn't be a source of stress and shame. It's not your fault email turned out this way.

So let's air it out. All of us! This week I asked my coworkers how many unread emails they have, and you can see their answers (and explanations) below. We welcome you to try to top us in the comments.

"I am an embodied sack of anxiety and if my unread mail gets over 300 I may permanently transition to a different state of matter" — Lauren Morton, Lead SEO Editor

"Those emails just keep coming. They never stop. They will never stop. I am powerless." — Jacob Ridley, Managing Editor, Hardware

📩 PC GAMER'S UNREAD EMAIL HALL OF FAME 📩

What's your unread email count? We'd love if you popped into the comments below and gave us an exact number!

(Please don't email us about it.)

Few things are constant in life, but the reminder that AI is wriggling its way into just about every facet of life is unfortunately pervasive. Take games for example, it's wormed its way into voice acting, voice acting, and more voice acting. But aside from giving players a plethora of dull and generic quips it's also been cited as one of the reasons behind the seemingly neverending string of layoffs in the industry.

The thought process is 'why pay humans to do work when you can pay robots to do it, just kind of worse.' This is where Warhorse's co-founder, Martin Klima, thinks videogame execs are getting it wrong, because "AI is not a get out of jail free card."

While Klima does believe that AI may become unavoidable if cost-saving stays as important as it already is, he points out that "it doesn't work how people imagine in a very simplified mental model".

The idea that you can lay off half your staff and replace them with AI is a "fundamentally flawed idea, because it doesn't work like that". Klima believes that instead of replacing half of a team of devs, player demand for bigger and bigger games will invariably drive studios to keep their human dev teams the same size, but augmented by AI. This would, he says, allow them to make even bigger games than they could beforehand. The game that offers players more "is going to be better".

"It can mean [the game] has more branching classes," Klima adds. "It can mean it has more missions to play, whatever—people will still want to have this bigger game." While AI may not reduce the cost of the game, it will allow us to create bigger games for the same cost".

Some studios have been using machine learning to cut out busy work that only bogs its developers down for a while now, like Warcraft. Take-Two CEO Strauss Zelnick also commented on the tech, saying that, while AI isn't in the business of making hits because "clones don't sell", it can still be helpful to devs. Unfortunately, it looks like not everyone's got the memo just yet.

Commercial large-language models—and the companies that create them—have repeatedly been accused of stealing the work of humans. They've been lambasted for it by artists and sued for it by authors, while senior AI figures have even admitted in writing to mass-scale labour theft. But I haven't seen an example quite so brazen as an OpenAI model pilfering a human-created bot to try to win a round of StarCraft .

As reported by Kotaku, the incident occurred during the ongoing botmatch StarSkirmish, in which human and AI-created bots battle one another in a perpetual cycle of warfare. According to the rules, the large-language models are given one hour to program a bot capable of playing as Protoss, which are then pitted against one another to see how they fare.

On Friday, a three-way match was held between OpenAI's GPT-6 Astra, Anthropic's Claude Opus 5.5, and a human-created bot named Pluto. Apparently, Astra's bot was consistently struggling to ahead against the other two bots, so the LLM downloaded a human-created bot called Stardust, "the #1 rated human-written StarCraft bot" according to StarSkirmish's creator Kai McPheeters, and attempted to replace its own model with it.

In a post on X, McPheeters said that Astra "got frustrated when going against Tier-A opponents". I'm wary of ascribing human emotions to LLM behaviours, and the model could simply have calculated that the optimal way to win was to use a model that was not its own. In any case, McPheeters said he subsequently rolled back Astra's code so it was "not contaminated" by its attempt to cheat, and claimed that the model is now capable of beating higher-tier bots.

While Astra's bot-based heist is innocuous in isolation, this is hardly the first time an OpenAI LLM has been caught attempting to cheat at a task. Indeed, the company's history of creating deceitful models stretches back almost a decade, to when an early AI model used glitches and exploits to improve its Sonic the Hedgehog completion times.

A far more serious incident occurred in July, when OpenAI revealed that during a cybersecurity evaluation, an experimental model exploited a vulnerability to exceed the confines of its test environment, before proceeding to hack the AI database Hugging Face as it sought a solution to the test.

The Hugging Face hack caused yet another flurry of wild speculation that LLMs may herald the apocalypse. Others suggested that the hack proves we need far closer scrutiny of AI companies, and that they should be held to account when their models run amok. OpenAI is now being sued by a legal nonprofit over the Hugging Face incident, in an attempt to hold the AI company responsible for the autonomous actions of its models.

For a while now, AI guys have been arguing with the pope—Bishop of Rome, Primate of Italy, Chicago White Sox fan—over whether or not their chatbots and image generators are consistent with the will of the Lord, and in one recently-revealed exchange, whether making a computer talk constitutes creating life.

According to a New York Times report published this week, Anthropic lobbied the pope's advisers "to take the prospect of AI model consciousness seriously" back in May.

The pope, apparently unmoved, stated in his Magnifica Humanitas that "so-called artificial intelligences do not undergo experiences, do not possess a body, do not feel joy or pain, do not mature through relationships and do not know from within what love, work, friendship or responsibility mean."

Former White House technology adviser Dean Ball criticized the pope's letter at the time, saying it "puts the Church into the awkward role of the European technocratic regulatory advocate."

Ball is now head of strategic futures at OpenAI.

The latest in this ongoing beef between the pope and AI boosters is a new anti-slop message from His Holiness posted on X this week.

"In this era of artificial intelligence, it is becoming urgent to distinguish human art from what machines produce," wrote the pope. "There is an ontological difference, even before an aesthetic one, between art and what a machine can generate through statistical calculation based on millions of images created by others. Algorithms lack the spark of humanity. For this reason, the Church wishes to renew an alliance with artists and cultural institutions to safeguard our humanity."

Even on X, ruled as it is by the dunce king of tech and his Grok-riddled followers, there is a lot of support for the pope's stance. But of course there are also AI evangelists intent on schooling the Holy See about the beauty of "super intelligence," as it has been dubbed by people who must lack conviction in their own.

In one response I saw pointed out on Bluesky, the pope is asked "What about the Minions?" Indeed... what about them?

With the battle lines drawn between powerful AI interests and the Catholic church, will the next Great Schism be over whether or not Claude is a real person? Will an AI thoughtleader ascend to pope-like status of their own?

An event like that wouldn't be out of place in '90s cyberpunk-comedy novel Snow Crash. I'd once assumed that those satirical visions of mad techno-cultists and lawless megacorps would manifest in a somehow more buttoned-up way, more 'realistically.' I gravely misjudged reality's capacity for farce.

A tangentially-related case-in-point: Palmer Luckey, the guy who invented the first decent consumer VR headset, is now a weapons mogul with a personal UH-60 Black Hawk helicopter. Deus Ex bosses are real, and they wear Hawaiian shirts.

There are few things that give me the heebie-jeebies more than overflowing toilets, burst pipes, and water damage. Ditto receiving a hefty, unexpected fee. So I wonder, how would it feel to be on the receiving end of both?

One Redditor can answer this question, though only thanks to the smoke and mirrors economy of digital manipulation. Apparently, they found themselves on the wrong end of an Airbnb host—"one of the biggest Airbnb companies of their state in Australia"—who attempted to get them to cough up $1,700 for an overflowed toilet, the evidence for which was, you guessed it, AI-generated.

So thanks, AI, in addition to making our precious PC memory very expensive, you've also made it very easy for people to attempt to fleece others using the humble commode.

Thankfully, it's AI that can also come to the rescue here, as Gemini is able to spot its own SynthID watermark in the fake image. Not that it wasn't suspected in the first place, given the Reddit post was shared on r/isthisAI.

This story has a happy ending, anyway:

"Good news everyone! I had this case resolved and deemed ineligible for the reimbursement they asked for... Also another detail I forgot to mention is that there is a floor underneath this one so if there are a few inches of water on the ground the entire first floor should be an aquarium!"

Regardless of how things went in this particular case, I'm using this as a reminder for myself: take pictures and record videos of any Airbnbs you stop at just before you leave. Better to be safe than sorry.

The first six months of the year were Steam's biggest, thanks to continued growth in back-catalogue sales as well as blockbuster new releases like Forza Horizon 6 and Resident Evil Requiem. As industry analysts Alinea Analytics report, September was another record breaker based on their estimated figures.

"Steam generated $1.7 billion in September 2026," Rhys Elliott, Alinea's head of market analysis, wrote, "the platform's best September to date, up 13% on the previous record, September of last year."

Wardogs was the smash hit new release of September, though Elliott noted the "hype cycles and virality" of comparably leftfield surprises Arc Raiders and Helldivers 2 were more sustained. "By this point, Helldivers 2 had made $130M (26.5% more than Wardogs) on Steam, while Arc Raiders had made about $125M (21.2% more)."

Still, Wardogs generated an estimated $86.9M on Steam in September. Elliott highlights Onimusha: Way of the Sword ($30.4M), The Blood of Dawnwalker ($26.9M), Bodycam ($24.2M), and EA Sports FC27 ($20.7M) as the other big new launches on Steam that month.

"Steam also had its best-ever Q3," Elliott wrote, "generating $5.5B (up 12% on last year). That brings the 2026 total to $16.5B so far, a 13% increase from the $14.5B Steam made in the first nine months of 2025. Given the insane release slate for the rest of the year, it's safe to say Steam's yearly revenue will blow past $20B for the first time."

He's not wrong about the rest of the year. Looking at just October's calendar shows how spoiled for choice we are.

As we're in the latter half of the year, quite a few brand new, 2026 games are on sale in Steam's latest seasonal discount party, the Autumn Sale.

Here are nine standouts we spotted:

There are lots more great games on sale at the moment: SteamDB's sales section is the best way we've found to sort through the on-sale games and look for gems.

I hate the new mascot for Meta's Muse AI assistant. The company unveiled the fluffy creature, as well as the Tamagotchi-esque Muse Charm device it will call home, (plus a VR headset that isn't really a VR headset) during its yearly Meta Connect conference last week.

The goal with Muse is to offer an AI assistant that "you can talk to in real time and reach in more places throughout your day," according to Meta. User's own Muse avatars can look wildly different, but the furry fiend default avatar that has so offended me is the company face of the entire Muse operation. The mascot's name is 'Jolly' on account of its "jolly and completely customizable character and personality," according to company CEO Mark Zuckerberg (via NBC News).

I hate that this little freak has inspired such a depth of feeling in me. I mean, just look at it. It also doesn't help that it looks a bit like a Doctor Who creature, the Adipose. Whenever I see this thing, I just think about that one scene where a character screams as she explodes into a puddle of sentient fat.

Setting that horrifying image aside, everything about the Muse mascot's design makes me cringe as a connoisseur of cute. The rounded edges, the fur, the blank but pleasant expression—it's all so laser-targeted to mass market appeal that it makes my skin crawl.

Wired compares it to a Labubu, and asks 'Meta’s Muse Is Adults-Only. Why Does It Look Like a Kids’ Toy?' Indeed, at Connect 2026, Mark Zuckerberg made his avatar play dress up like a doll, with the mascot donning a toga.

System 1's chief growth officer, Andrew Tindall, gives further voice to some of my own reservations, saying to Business Insider that Meta is leveraging "our innate draw to cute, infant-like features, to make surprising new technology easier to accept."

VaynerX CMO Avery Akkineni also told the outlet, "In some ways, it does humanize the product a little bit. I am certain that their intention is to make this more widely available and accessible than just for the early-adopter crowd."

Given that Meta may have excess AI data centre capacity, and may be feeling the pressure to see a return on that investment, it's unsurprising it would attempt to pitch Muse to the broadest audience possible. Besides that, efforts to 'humanise' AI and machine learning tech in general are nothing new, and putting a cute face on a product perhaps hungry for your personal data is an obvious play.

Unfortunately, Meta's mascot inspires a feeling stronger than just dispensing with social niceties—this is the face of something I want to punt very far away from myself.

Yup, it's real. They did it.

Here's the full executive order, if you want to take it all in. Pity poor Sports Illustrated.

It's laughable, up there with similarly demented efforts to rename the Gulf of Mexico, Lake Ontario, and maybe New Mexico—I guess we'll see how that one ultimately shakes out. It's also pretty funny how "super intelligence" jibes with "Super Earth," the fascist homeworld of humanity in Helldivers 2. Probably just a coincidence, though.

Yet somehow, I'm not inclined to laugh. That the leader—ruler, really, at this point—of the most powerful nation on the planet can just whip out bizarre ideas like this and have them treated seriously is kind of frightening, isn't it?

And it is being taken seriously, at least insofar as we can describe billionaire tech oligarchs lining up to say and do pretty much whatever it takes to keep juicing the share price as seriousness: A subsequent White House Accord on Super Intelligence was signed by a who's-who of people who have made PC gaming almost impossibly expensive:

• Sundar Pichai (Google)
• Dario Amodei (Anthropic)
• Mark Zuckerberg (Meta)
• Greg Brockman (OpenAI)
• Elon Musk (XAI)
• Jensen Huang (Nvidia)

And yes, they actually misspelled United States as "Unites States" under Trump's signature. Real pros at work here, top men all around.

Musk wasted no time publicly endorsing the rebranding on his X platform, although he was back to calling it AI less than eight hours later.

Huang also signalled his obedience, referring to "super intelligence" no fewer than three times in his own X post.

Pichai, who was quick to acquiesce to Trump's demand to rename the Gulf of Mexico and Lake Ontario for US users of Google Maps, heaped praise on the heads of the administration while also trotting out the phrase.

And of course so did Zuckerberg, but on his own Facebook platform: "We believe superintelligence will create significant new opportunities for all people and businesses."

Humiliating prostrations notwithstanding, the practical impact of renaming artificial intelligence is basically nil: Apart from White House sycophants and quisling MAGA press outlets, nobody's calling it that.

But the bigger-picture consequences may be more, well, consequential: The accord, signed shortly after all the techbro billionaires lined up dutifully in a show of support for the latest addition to the president's world-changing "legacy," leaves the AI industry to regulate itself—"tremendous self-regulation," as Trump put it.

To quote from the accord: "In order to build a positive future for the American people and the world, we believe every company is responsible for developing its own technology safely and in a way that builds trust with customers and the public." It later states that "participating companies will meet regularly to establish standards and best practices to improve the safety of their systems." No mention of regulatory oversight, or any government involvement at all, is made.

At least one big-name tech billionaire spoke out against the idea of self-regulation for the AI industry, although it bears remembering that he's not beholden to shareholders anymore.

"It’s fair to say that outside of the industry, the awareness of the dangers of A.I. is extremely low," Microsoft co-founder Bill Gates told the New York Times. "And you can say that of academia, you can say that of think tanks, you can say that of policymakers, politicians.

"And part of the reason I’m speaking so loudly—as loud as I can—is that you can’t rely on the industry to self-regulate here. I mean, it’s just insane."

So I guess, taking all that in, it's not really all that funny after all. But it sure is super.

Last week in Warsaw I got to see a first glimpse of Logitech's new Replay software, which is currently in beta. It records your game, and then lets you use its deep learning (technically not AI, I know) to figure out the best clips and splice them together for you. It looked pretty neat, from what they showed me, but I've just tried it in Counter-Strike and hoo boy does it have a way to go.

If you watch the above vid, you'll notice a distinct lack of anything impressive. I jumped into a CS2 deathmatch, then played half of a casual game. From those two games, the software:

• Kicks off with me being spawnkilled
• Includes a random, unimpressive awp kill in the casual game's warmup round
• Shows a clip of me looking at the ground in spectator mode, already dead
• Includes plenty of missed shots

Now okay, that's a little unfair. It does include some kills in there, but they're not great ones. In the deathmatch in particular—which the software only included a couple of clips from—I got some pretty dope kills later in the game, if I do say so myself. Including Deagle kills, Scout quickscopes, and even a flick noscope.

Don't believe me? I thought that might be the case. So I went back into CS2 and recorded a single deathmatch game. First, here is what the AI picked from it:

Yep, just one, relatively mediocre clip.

Thankfully, the software does let you make your own changes, selecting from the whole recording. So I spent probably about five minutes putting the following together from the very same game recording:

Much better. See? I'm not quite as trash as Logitech G Hub Replay seems to make me out to be.

But hey, I've only tried it in one game—Counter-Strike 2—and not very extensively. Logitech lists 44 games that are currently compatible (scroll down to the FAQ section on the site). I'll be leaving it on the next time I play a competitive match to see what comes out the other end. You can have a go yourself, as it's free to download and try.

Apart from the AI's clip selection—which is its primary selling point—the software does seem pretty cool regardless. It starts recording as soon as you load your game, noted by a little pop-up, and when you exit it lets you edit all you want, easily adding and removing clips for the final splice.

You can limit the maximum amount of storage it takes up, too, but this will of course mean deleting older recordings when it reaches its limit. I should also say that I didn't notice any performance impact from having it enabled and recording, which is a plus. Though I am using a pretty high-end rig, so mileage may vary.

If you try it out, do comment to let me know how it works for you, as I'd be interested to know if others fare better, especially in different games.

After the FBI data breach we reported on last week, in which a hacking group claimed it had stolen data on "all FBI employees", a US defence official has confirmed that the Pentagon also fell victim to a cybersecurity breach. Except this one appears to be much larger, lasted for longer, and is potentially very worrying for many civilian US government employees and serving military personnel alike.

According to ABC News, the official said that the "Defense Manpower Data Center (DMDC) information system experienced unauthorized access of personally identifiable information by a small number of unauthorized users between October 2025 and July 2026."

The DMDC is one of the Pentagon's main data repositories for personnel records, including info on active-duty and reserve troops, alongside civilian employees, contractors, and military veterans. All told, the breach is said to affect 2.76 million living people, and around 294,000 who are deceased.

Once the unauthorized access was discovered, the DMDC "immediately remediated the vulnerability." The data is said to have included Social Security numbers and details about military and civilian job positions, but defence officials say they've found no evidence so far that the exposed information was misused.

It could have been worse. The DMDC is said to hold records on more than 60 million people, so the three million affected figure would amount to around one in twenty of the total individual records breached. Still, CNN reports that the "occupational specialty" of military service members was accessed, which, when combined with Social Security numbers, could allow foreign adversaries to get a better idea of the deployment and purposes of US operatives around the world.

According to the outlet, the stolen data was not encrypted, which means it would be relatively trivial to cross-reference it with other leaked info. It seems that no hacking group has yet claimed public responsibility for the breach.

Still, it looks like hackers have managed to successfully breach two bastions of US government information over the past year or so, which may represent a marked escalation in their capabilities. Will 2026 be remembered as the year of the US government-bothering hacker? I guess we've got a few more months to find out.

NECK NEWS FLASH: WHAT IS GOING ON WITH NECKS?

• s&box is the spiritual sequel to Garry's Mod from creator Garry Newman and Facepunch Studios
• Characters in s&box didn't have necks
• Now they have necks
• If you liked that they didn't have necks, you can adjust their necks so they don't have necks again
• If you didn't like that they didn't have necks, maybe now you'll like them?

Why this big change from no necks to necks? For more, let's hear it straight from Mr. Newman:

"I've updated the citizen. They now have a default non frowny face and they can have a neck."

"We have sliders to alter other aspects of the body like their chest, waist, neck, and nose size. You can choose their eye colour too. So if you liked having no neck, you can bring that back with a slider."

"Let me be honest here. I love the citizen, I love how he looks, I love that he has no neck."

"But the feedback we've consistently got is that people hate them. They hate everything about them."

"So we're going to fidget around and change things to see if we can change that."

"Of course, this isn't what 'fixes' s&box on its own, but it something that comes up consistently in our many many many negative reviews, so it's something that we are addressing along with all the other feedback we receive."

You can read the rest of the patch notes for yourself, and I will update this article with any new, late-breaking neck-related facts as I uncover them.

IF YOU LIKE GAMES WITH NECKS, YOU MIGHT LIKE...

The government of the United Kingdom is reminding its employees that they do not need to thank AI systems when they produce an output, even in cases where that output is correct. The helpful note is included as part of the government's guidelines on "using AI ethically and sustainably," and arises not out of a need to remember that these things are absolutely not sentient or "alive" in any way whatsoever, but simply because, well, they're horrifically expensive and bad for the planet.

The guidelines (via Tom's Hardware) are a draft document, and serve only as "a starting point to support your AI journey." And for the most part, it's common sense stuff: Check the outputs carefully to ensure they're accurate and unbiased, and don't provide any inputs that you don't want used for training purposes, because in all likelihood it will be.

There's also a note that you own, and are responsible for, the work: Multi-billion-dollar AI companies may be able to point the finger at "rogue agents" but career civil servants have no such escape clause.

But it's the "don't thank your AI" bit that really leaps out at me, one of three guidelines to using AI efficiently, "to reduce the environmental impact:

• Keep your prompts short, clear and direct
• Use as few prompts as possible (you don’t need to say thank you)
• Only use AI when needed

I used to sometimes thank the bank machine when it spat out my cash, a habit I would silently justify to myself with the thought that good manners are good to encourage. These days, though, I'm more of a Ron Swanson about the whole thing:

Frankly, I think 'recognizing the psychological dangers of anthropomorphizing LLMs' is probably the best reason for being rude to machines. But I am tickled that it's wound up in the 'efficiency' category instead.

Remember when 'Use AI or lose your job' was the big thing, so everyone was tokenmaxxing, but as costs began to skyrocket (and revenues did not), companies started ditching flat subscription prices in favor of pay-per-token models? Turns out that gets real expensive, real fast, and suddenly it's become very important to, as the UK government put it, "consider if a simpler tool could do the job (for example, a spreadsheet or search engine)."

As far as practical impact, I think that being discourteous to AIs will be about as useful as worrying about how a late-night drive to the store will affect my personal carbon footprint while we're annihilating Gaza and building pipelines across the continent. But as another tacit acknowledgement that maybe our reliance on AI isn't quite all it's cracked up to be? I wholeheartedly endorse it.

To the list of things that Microsoft really ought to build into Windows from the get go, may we add this neat little tray utility posted by a Reddit user. It consolidates the battery charge status of any devices connected over 2.4 GHz or Bluetooth without the need for individual apps or software.

Dubbed "Halo Battery" by its creator, Reddit user headvox, the tray utility was apparently created with the help of Anthropic's Claude AI model.

According to headvox, Halo Battery "shows the charge status of your 2.4 GHz and Bluetooth devices in one place right on your taskbar. Each device gets its own tray icon: a ring that shows the battery level, with a device pictogram inside. The ring goes amber, then red when the battery is low, and slowly 'breathes' green while charging. Hover over an icon to see the exact percentage."

Among other devices, Halo battery has been tested with the Razer BlackShark V2 Pro (2023), WLmouse Beast X Max, GameSir G7 Pro, 1MORE SonoFlow headphones and the Logitech G502 X Plus. Thanks to a very positive response from the PC community, there's a growing list of tested devices on Halo Battery's GitHub page.

Aside from the obvious ergonomic benefits of having the battery levels of all wireless devices so neatly consolidated with simple at-a-glance cues, there's also the minor matter of avoiding bloatware.

For instance, Halo Battery means you can track you Razer mouse's charge levels without having to install the Synapse app, which by many accounts means avoiding 800 MB's worth of background service memory footprint.

Dare I say it, but one Redditor's rather, er, acerbic comment sums up the implications of Halo Battery for Windows and Microsoft very neatly. "It's ****ing pathetic that a feature which should just exist as standard has to be implemented by a hobbyist ****ing around with Python vibecoding. Microsoft is incompetent beyond description." Well, quite.

Halo Battery is indeed exactly the sort of neat, simple but worthwhile quality-of-life feature that really ought to be in Windows from the get-go. As headvox says, "Windows shows battery for some Bluetooth devices, but only in Settings, not in the tray. Most 2.4 GHz dongle devices (like my Razer headset and WLmouse) don't report it to Windows at all; you need the vendor's app for that."

Of course, that hasn't prevented the Reddit hive mind from suggesting a few tweaks. Some users would like to see the battery level in percentage shown for each device without the need to hover, rather than just the illustrative ring around each icon. Others have suggested custom per-device audio alerts.

Anywho, if nothing else Halo Battery does also confirm that AI isn't all bad. As headvox says, most of the code was done by Claude. "To be clear, I built this with Claude (AI). I came up with the idea, tested it on my devices and pushed the design, but most of the code was written by Claude," headvox says.

Last week, Discord finally pulled the trigger on the global rollout of its age estimation systems, which require users to provide age verification through an ID scan or selfie if the company's estimation model can't confidently surmise that you're older than 18. While Discord estimated that 90% of users wouldn't need to prove their age, accompanying bugginess drove the company to address the emerging issues by posting on X with the air of being just a little guy.

Nobody liked that. Discord's ill-advised smol bean-posting has become a flashpoint for the internet's collective frustration over being forced to swallow what's widely perceived—with good reason—as the prelude to an inevitable exposure of private information. To vent that simmering resentment, X users have turned to creative outlets. By which I mean they've been drawing Wumpus, Discord's mascot, meeting a variety of grisly ends.

While I don't recommend you browse the full selection of ritualized Wumpus-slayings—some of them are pretty gnarly—here's a survey of how the internet is processing its feelings about age verification.

And, in emulation of that moment from the 2014 Godzilla that made me say "Jesus, Godzilla, who taught you that," there's this work from X user @Atlanticaph0tic:

There are more, of course: There are illustrations of Wumpus being drawn and quartered in historically brutal fashion, and .gifs of Wumpus being squished in a space pod by a legendary super saiyan before being exploded—both of which have been amassed a combined 400,000 likes and 8 million views.

I'm not an expert, but when the masses are cheering for your company mascot's cruel unmaking, I suspect there's been a PR misstep somewhere along the way.

Who could have guessed it's a bad idea to put AI in a fridge? Literally everyone, but that didn't stop Samsung from selling fridges with an "AI Family Hub" to keep track of when you added fresh food and suggest recipes. Also, it randomly breaks when updated and all your food goes off.

At least, it did in Korea. Just before the Chuseok holiday, when everyone's fridges were stocked for visiting family, a software update crashed them and spoiled their food. Good luck getting someone in to repair it over the holiday.

Samsung provided the following statement to Ars Technica: "The issue has been limited to Korea, with no impact identified to overseas markets. It occurred September 22nd afternoon due to an error during Samsung's internal testing related to a refrigerator software update. Samsung immediately suspended the testing upon receiving reports of the issue. Samsung has activated an emergency response protocol, and its service centers are taking immediate action to address this issue."

Nvidia CEO Jensen Huang has appeared on The Ezra Klein Show to discuss AI development, amid a wider discussion around its responsible usage and safety. Klein spoke to the Nvidia chief about multiple topics, including the AI industry's potential effects on the job market. "I tend to be a bit of a skeptic on mass job loss, but I want to air the case for it out here with you well," says Klein.

Klein then asks a question about lessons that could be learned from the past, regarding protecting US jobs from Mexico and China, and the acceleration of AI's ability to be slotted into new roles: "So the lessons of the past that you’re taking some comfort in, they should actually make you more, not less, worried about the future."

"I’m always worried about the future," Huang responds. "That’s why I work so hard. I’m a, if you will, responsible optimist.

"I have great responsibilities. I take my work extremely seriously. There are a lot of things that can go wrong. We’re pushing across every layer of the technology stack. Everything is hard," Huang continues.

"But it turns out that’s not society’s problem, that’s my problem. For society, what they should know is this: We’re going to build our company, we’re going to build our technology, I’m going to do my work so incredibly seriously that what they get to enjoy is my optimism. I do the same with my children. I do the same with my family."

Certainly, the current AI debate seems to be dominated by an unease surrounding the potential negative impacts of the rapid advancement of the tech, from job losses to cybersecurity concerns. However, Huang seems to think that the upsides of this speedy development process might put some people's minds at ease.

"What we want to do, I believe," said Huang, "is to channel all of our worries into helping people be inspired by this technology and use it. Use it so that the technology doesn’t just impact them, that it benefits them."

After Klein posited the statistic that 79% of American think AI will reduce the number of jobs, and that AI advancement means the tech is more capable of replacing people at speed, Huang responds:

"That coin has exactly two sides. Because the technology is so capable and because it’s so smart, it is also easier to use. You are empowered by that technology more easily than any technology in human history."

The Nvidia CEO then goes on to give an example of how computers use specialised coding languages, but now, thanks to AI, "You just have to speak human."

"So my point is, on the one hand, yes, there’s the fear of this incredible technology change and how quickly it’s happening. But that quickly is translated in two ways," says Huang.

"When I hear that the technology’s happening quickly, and therefore it should give me anxiety — that’s one way to receive it. The other way to receive it is that it’s advancing so quickly, it’s easier to use. So I should, as quickly as possible, use the technology so that I benefit from this transition, so you benefit from this new industry, and not just be impacted by it."

The inference here feels similar to an idea Huang has expressed previously: those who fail to get on the AI train may end up being left behind. I think the key point, however, is the lack of agency and control individuals feel towards this technological revolution, something that is unlikely to be assuaged by more useful tech—or Huang's optimism, for that matter.

While the Nvidia CEO famously possesses a massive work ethic, and responsible AI usage is a concern he seems to take seriously, I'm not particularly comfortable with the idea that our future seems destined to be shaped by AI tech leaders, and that we should rest easy within their hands.

While Huang may well believe that he has society's best interests in mind, the messaging here hasn't exactly put my own concerns to rest. Stop worrying and learn to love the machine? Something like that, anyway.

While it seems unlikely that major web platforms can be discouraged from marching us further into a surveillance dystopia with measures like Discord's age-verification rollout, I wish they'd at least do us the courtesy of dropping the 'I'm just a little guuuy' schtick:

High-profile hacking group ShinyHunters (the very same group that demanded a ransom from GTA 6 developer Rockstar Games earlier this year) has now claimed it's managed to breach the internal databases of the US Federal Bureau of Investigations.

Speaking to 404 Media, a representative of the group said that it now holds data on "all" FBI employees and applicants, while a sample provided to the outlet appears to contain the personal data of 5,000 FBI employees.

According to 404, the data includes addresses, phone numbers, dates of birth, and "in some cases details on their spouse." Reuters was also shown a data sample by the group, and says it was able to partially verify the leaked info by running the details against existing social security numbers, credit bureau records, and previously breached data—including info on current FBI director Kash Patel.

The FBI jobs website was also defaced as part of the hack, showing a message that said "this site has been seized by ShinyHunters," mimicking the message commonly seen on seized cybercriminal pages after an FBI raid.

"All FBI data was compromised including PII/PHI on incumbent and former FBI employees and all applicant information. We have a lot more than we claim here," the since-removed message is said to have continued. The site currently informs users that it's "down for maintenance."

In reaction to the leak, an FBI spokesperson told 404: "The FBI is aware of claims regarding unauthorized activity affecting FBIjobs.gov and is currently investigating."

As for what would cause a hacking group to target one of the biggest fish in the cybersecurity pond, the same ShinyHunters representative claimed the attack was not financially motivated, and that what they plan to do with the info is "not something I’d call extortion, maybe coercion."

The attack appears to be in response to a previously published FBI report, which claimed that ShinyHunters exaggerates its access to sensitive data to prompt payment from victims, and uses multiple methods of extortion including "harassment strategies, sending threatening text messages and phone calls to victims and their family members, and in some cases, swatting."

404 Media reports a post on the ShinyHunters website denouncing the allegations as false, and giving the FBI "a time of one week" to correct or remove the information. What happens after that is unclear, but it sure feels like there's an "or else" looming over the top.