AI

Though OpenAI didn’t give its internal research model access to funds on the network, the AI system instead found a way to access another account with test crypto, using it to help complete its task, according to OpenAI.

Beijing is reportedly considering adding 23 AI technologies to its export-control list, including chip designs, agents, model weights, and source code. China has been steadily gaining on the US with recent releases like Kimi K3, and has been the target of Western AI-related export controls amid growing competition to maintain control over the strategic technology.

In a noncommittal X post, Elon Musk said SpaceX will “probably” follow in Anthropic’s footsteps and allow external evaluators access to its AI systems. The tentative pledge comes amid growing concerns about the risks of frontier AI models and criticism that voluntary measures leave companies largely policing themselves.

The drills were confirmed by Japan’s finance minister on Thursday. States and organizations have grown increasingly worried about AI’s ability to supercharge hacking, particularly after months of impressive releases and rogue agent attacks. Experts, including the heads of the Bank of England and International Monetary Fund, have stressed the need to guard financial institutions.

“How do we build deeply personal experiences leveraging AI that help us all be more creative, be more productive, do whatever it is that we want to do,” Ternus says in an interview with TIME.

Meta just shipped a pair of smart glasses with no camera — a change that both David and The Verge’s Victoria Song are excited about. But these glasses, like the new product from Amazon Bee and seemingly everything else now, are still ready and willing to record the world around you at all times. What do we do in a world where everything, everywhere, might be recording? And can tech companies figure out a way to make us feel better about it?

It also demanded documents related to the AI companies’ safety practices, consumer privacy policies, and marketing materials, as part of “a probe focused on potential consumer protection violations,” Axios reports. Its sources also say the FTC “plans to compel top AI executives to testify about their products.”

The “Ask YouTube” feature, which pulls uses AI to generate a search results page populated with text and suggested videos to watch, popped up on mobile as an experiment in August. I tested the web version earlier this year, and in a few minutes using the mobile version, it’s pretty similar.

Six employees at the AI safety organization METR will gain access to hardware and internal tools “roughly on par” with what Anthropic’s own risk assessment team sees. While Anthropic says METR will publish reports about Anthropic’s risk levels, incidents, and practices, it can still instruct the safety evaluator to redact certain information, such as the inner workings of its models or customer data.

In a deep-dive profile of Marc Andreessen’s petty grievance bullshit, Bloomberg Businessweek gently points out that he is not exactly a member of the reality-based community. Since covid, that’s gotten worse, as he and his wife have cocooned themselves away from ever encountering a normal person.

The website decomp.dev published a big statement following the recent influx of AI-assisted decompilations of games:

The funding is part of a $15 million initiative led by the Simons Foundation, with Anthropic also contributing. It follows months of upheaval in the field as AI companies race to solve longstanding problems and comes just over a week after OpenAI dumped hundreds of preprints on the research community.

He claims in a post on X that it’s “never happened yet.” Musk didn’t go into details, but he’s made other promises before, like saying he would pay legal bills for people fired unfairly over tweets, or his many, many claims about robotaxis and Tesla’s new Roadster.

Former Boston Dynamics CTO Aaron Saunders, who joined Google in 2025, announced the project on LinkedIn. “We’re creating a first-class lab to house this team and take robotics to the next level,” he said. It comes as Google pushes deeper into robotics, including with Gemini Robotics AI models and absorbing Alphabet robotics moonshot Intrinsic.

The chip manufacturer reported a net income of NT$763 billion (about $24 billion) for the quarter, up 69 percent compared to last year, and it’s predicting it’ll do even better in Q4. It’s a good time to make computer chips.

Shortly after US District Judge Edward Davila said he was “highly likely” to grant Apple’s injunction, OpenAI voluntarily conceded instead. It agreed not to use Apple’s proprietary information for product development while the rest of the lawsuit plays out, and will hand over forensic images of the drives of 14 employees who defected from Apple.

BofA CEO Brian Moynihan doesn’t want people turning to the banks when an agent screws up a transaction. He says the bank is working with the AI industry to argue for human payment authorization, at least until the risk of error is mitigated by “further development of the capabilities” of agents.

As spotted by Kotaku, some people are posting videos showing strange visual issues with the updated launcher that agents wrote code for. Right now, the launcher is in preview.

Yesterday, the DOJ and OpenAI revealed that Abdulrahman Alfozan, a 35-year-old Saudi national and former Facebook employee working for OpenAI, has been arrested and charged with installing a recording device behind a TV in an executive conference room.

The launch will include some of Apple’s earlier Apple Intelligence features, but not the just-launched Siri AI, according to Bloomberg. The update is set to release in December.

Here’s a tidbit in SpaceXSI’s announcement post for Grok Voice 3.0 today:

The company’s unsecured debt due in 2032 is now trading below 80 cents on the dollar, Bloomberg reports. “Yikes” is right, Elon — if CoreWeave crumbles, other debt-laden neoclouds might be in trouble, too.

A man named Mike Smith was recently sentenced to 18 months in prison, and a fine in excess of $8 million, for creating an AI system that faked billions of streams and earned him millions in royalties. Wired’s Kate Knibbs explains how Smith pulled it off, how he got caught, and what this case means for the internet’s growing slop problems.