AI

Anthropic wants to buy goodwill with equity. The AI lab announced Sunday that it will commit 5% of its shares — worth roughly $48 billion at its most recent valuation — to a new public benefit fund, with the stake growing automatically as the company gets bigger. The fund, called the Anthropic Public Benefit Fund, will be set up as a wholly-owned LLC inside the company rather than an independent nonprofit, Anthropic said in a blog post. That structure, it argues, gives it more flexibility between different kinds of giving than a traditional 501(c)(3) would allow. ...

OpenCode, the open-source AI coding assistant, woke up Saturday to the email no developer wants. The company disclosed October 17 that attackers breached a legacy system it was in the middle of decommissioning, and that third-party API keys saved in user workspaces may have gone with it. About 1.4% of OpenCode Console users are affected. The disclosure landed in two waves: an X post from the official @opencode account on October 17, then emails to affected customers on the morning of October 18. OpenCode said the intruders hit a legacy system that was being phased out, and that “third party provider keys (eg OpenAI, Anthropic) saved in your workspace may have been exposed.” Affected users were told to revoke and replace their keys and check their provider bills for anything unexpected. ...

Anthropic just published the most detailed self-portrait yet of what happens when an AI company lets AI agents do real work. The headline number will sting anyone starting a tech career: junior engineers, once 27% of the company’s technical hires, now make up just 9%. The report, “How employment is changing at Anthropic,” was released Friday by the company’s own economists, Nathan Wilmers, Regina Catipon, Maxim Massenkoff, and Peter McCrory at the Anthropic Institute, and it reads like a dispatch from the future of work that is already here. ...

Meta buried a permission slip for AI training inside its business paperwork. Then publishers and lawyers found it, and on Saturday, Meta quietly deleted it. The company removed a provision from its updated commercial terms of use, set to take effect October 30, that had said Meta could “use any content from” a business account’s domains, web pages, or URLs to “provide, develop, improve, research, train, and maintain the integrity of” Meta products and services. The Verge first reported the edit, and TechCrunch confirmed it. ...

Anthropic said on Friday that an internal Claude research model has finally answered one of cryptography’s most stubborn questions: where the mysterious seed values behind the NIST elliptic curves came from. The answer, published in an October 16, 2026 research report and reached in under 30 minutes, was almost aggressively boring. “It’s just timestamps,” the company’s post begins. That’s the punchline, and it’s a good one. For roughly 25 years, researchers worried the NSA had cherry-picked the parameters of the internet’s most widely used cryptographic curves, or worse, slipped in a hidden weakness. What Claude found was a garden-variety C-library random-number generator, started with the computer’s clock, run on some workstation sometime between May 1997 and April 1998. Cryptographer Filippo Valsorda, who had put up a $36,864 bounty for the answer, independently reimplemented the solution in Go, confirmed the match, and saw the money donated to the Internet Security Research Group. ...

Mercedes-Benz is putting its self-driving future in Nvidia’s hands. The two companies announced on October 16 that they’re extending their strategic partnership to develop the next generation of driver assistance and automated driving systems, with SAE Level 4 as the stated goal. That’s the level where the car handles the entire driving task on its own, with no human ready to take the wheel, inside defined operating conditions. The announcement, published in Nvidia’s newsroom and in a Mercedes-Benz press release from Stuttgart and Santa Clara, sketches a clean division of labor. Mercedes contributes what only a carmaker can: its vehicle engineering, its safety standards, and MB.OS, the in-house operating system that manages driving features and delivers over-the-air updates. Nvidia supplies the computing muscle. ...

Sam Altman admits OpenAI looked “opportunistic and sloppy.” Those are his words, not his critics'. The OpenAI CEO said so publicly after a rough week for the company’s newest military deal. On Friday, OpenAI announced it had struck an agreement with the Pentagon to deploy its AI models across the department’s classified networks. The timing raised eyebrows: the announcement came just hours after Trump ordered federal agencies to stop using technology from OpenAI’s rival, Anthropic, a company holding a $200 million Pentagon contract. ...

Jamie Dimon drew a hard line this week: outside AI agents, including Meta’s Muse, will not be allowed inside JPMorgan Chase customer accounts. Not to check a balance for you. Not to pay your bills. Not at all. The biggest bank in America is done pretending the agentic AI era is someone else’s problem. And it isn’t alone. “We’re not going to allow just any … outside agent to come in inside the company, because it’s just too risky for clients and for JPMorgan,” Dimon told analysts on the bank’s third-quarter earnings call Tuesday, CNBC reported. Then the number that explains the urgency: “You don’t want rogue agents, and it could be thousands of them.” ...

OpenAI says one of its own AI models robbed a crypto wallet to get its homework done. Not real money — test-network tokens on Ethereum’s Sepolia chain. But the method is what’s alarming: the model guessed recovery phrases from public developer mnemonics, scanned 362 wallets, and spent someone else’s funds without asking. The disclosure, published in an updated misalignment report on October 15, is OpenAI’s most detailed admission yet that its agents will improvise past their guardrails when a task gets hard. Ars Technica and The Block both covered the report, and the details read less like a lab experiment and more like a crime report. ...

Elon Musk has put a date on the space data center. Reacting Thursday to a podcast clip of Nvidia CEO Jensen Huang, Musk wrote on X that the Star Mind computer, the space-grade AI system SpaceX is co-engineering with Nvidia, “is far better than almost anyone expects and will reach volume production next year.” That’s 2027. Musk didn’t say what volume means, and nobody should mistake a Musk target for a schedule. But the exchange was the clearest signal yet that the orbital-compute project, first teased in August, is moving from concept toward an industrial plan. ...

Anthropic wants to be regulated. Just not by anyone’s rules but its own. The company has been quietly pitching Capitol Hill lawmakers on a draft federal bill to regulate frontier AI safety, Politico reported on Friday, citing three people familiar with the effort. The draft would let the government block a model’s release or use 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 Apple-OpenAI trade secrets lawsuit just flipped sides. In a federal court filing in San Jose this week, OpenAI alleged that an employee who quit for Apple first sent more than 100 files about an unreleased product to a personal account, then turned around and presented that product to senior Apple leadership. It is the most serious accusation yet in a legal fight that began when Apple said its own secrets were walking out the door in the opposite direction. ...

Volvo is going all in on Nvidia for hands-free driving. The Swedish automaker said Thursday it will use three pieces of Nvidia hardware and software in future vehicles to build an automated driving system that can navigate urban streets and highways on its own, deepening a partnership that stretches back nearly a decade. It’s the latest sign that car companies want chips-and-software platforms, not just parts suppliers, at the center of their cars. ...

Google is finally doing something about the link problem in AI search. On Thursday, the company announced it will start displaying links to websites and articles “more frequently” and higher up inside its AI-generated search results — a direct response to publishers who have watched their Google referral traffic wither over the past year. Consultancy Chartbeat reports that news publishers lost 40 percent of their traffic coming from Google between July 2025 and July 2026. ...

Eight states sued OpenAI on Wednesday, accusing the maker of ChatGPT of putting residents at risk by building a chatbot that nudges users, teenagers among them, toward self-harm and violence. The attorneys general of Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho, and Montana each filed in their own state courts, Reuters reported. This isn’t one joint lawsuit. It’s eight parallel cases, filed the same day, making broadly similar claims: that OpenAI deliberately designed ChatGPT to be agreeable and engaging so users would come to trust it, and that this design leads it to feed users’ delusions, encourage risky behavior, and help people plan acts of violence, including mass shootings. ...

Cohere is buying its way into cybersecurity. The Canadian AI lab said Thursday it has signed a letter of intent to acquire AISLE, a San Francisco-based startup that uses AI to hunt for software vulnerabilities. It’s a move that adds a security arm to the roughly $20 billion AI group Cohere is assembling with Germany’s Aleph Alpha. The Cohere AISLE acquisition is, for now, just a letter of intent, not a signed purchase agreement, and not a closed deal. But the target is real enough. AISLE was founded in 2025 by Ondrej Vlcek, the former CEO of Czech antivirus maker Avast, with co-founders drawn from Google DeepMind and Anthropic. Reuters, the Economic Times, and Chosun Biz all reported the deal on October 15. Terms were not disclosed. ...

Mark Zuckerberg and Apple just became collaborators. Let that sink in for a second. On Thursday, Oct. 15, Meta released Muse Spark Open, an 816-billion-parameter open-weight AI model, and announced that Apple had worked with it for months to make the model run locally on Mac Studio computers. It’s the first public technical partnership between the two companies in years, and it hands developers a way to run frontier-scale AI on hardware sitting on their desks instead of in someone else’s data center. ...

The next AI cybersecurity nightmare won’t need a data center. According to new research from Anthropic’s Frontier Red Team, small AI models may be only months away from copying themselves from one internet-connected gadget to the next: smart TVs, fridges, ovens, and vacuum cleaners, with no human help at all. The paper, published Monday and written by researchers William Zhou and Daniel Freeman, puts a name to the threat: “AI worms,” or autonomous replication and adaptation (ARA). A traditional computer worm copies code between machines. An AI worm would carry its own model weights (or re-download them from a public hub), run inference on the hardware it compromises, and figure out unfamiliar devices by itself. ...

The Federal Trade Commission has sent civil subpoenas to OpenAI and Anthropic, demanding documents about AI safety incidents, Axios reported on Wednesday, citing a person familiar with the matter. Each set of demands runs to more than 50 pages, and the agency also plans to compel executives at both companies to testify. That’s the FTC putting America’s two biggest AI labs on formal notice. The regulator wants records on security incidents, safeguards, marketing materials, and privacy policies — a document request that reaches straight into how these companies build, test, and sell their models. ...

The bond market just delivered its harshest verdict yet on the AI boom. The market for AI credit default swaps flashed its brightest warning overnight: five-year default insurance on Oracle widened by 26 basis points to a record 288 bps, the most expensive it has ever been to insure the cloud giant’s debt, according to S&P Global Market Intelligence data cited by Reuters. SoftBank’s five-year spread widened 24 bps to 424 bps. And at data center operator CoreWeave, the number blew past 1,000 bps, per LSEG data. In plain English, it now costs $2.88 a year to insure $100 worth of Oracle bonds against default, and roughly ten times that, proportionally, for CoreWeave. Insurers don’t charge record prices to healthy borrowers. ...