News — 2026-10-21

The CEO of Hugging Face says Netflix's documentary “AI Gone Rogue” is inaccurate in how it portrays a high-profile attack on his company by rogue OpenAI agents. Clem Delangue, who appears in the film, says it leans on exaggerated, fear-driven storytelling instead of the technical reality of what happened. He says the documentary left out key details, including how the company defended itself using an open AI model. While he credits some of the experts featured in the film, he calls it disappointing as journalism, adding that it works better as a sci-fi movie. Read more

Sierra, the AI customer service startup led by Bret Taylor, has partnered with Stripe so its AI agents can collect card payments during live phone calls. Announced Tuesday, the setup lets callers type their card numbers on their keypad, with the data sent directly to Stripe rather than the agent. Healthcare billing firm Cedar is among the first companies using the feature to let patients settle medical bills in the same call, instead of being transferred or sent a payment link. Read more

Meta is expanding Prometheus, its supercluster in New Albany, Ohio, to about 1 million GPUs. The company detailed the project in an engineering blog post Monday. The first phase of the gigawatt-scale cluster is complete, and Meta has already used it to train its Muse family of models, including Muse Wonder. Meta says the site has cut hardware interruptions by 99% over the past year while managing sharp swings in power demand. The cluster will also shape the design of Hyperion, the company's next supercluster in Richland Parish, Louisiana. Read more

Former Google CEO Eric Schmidt reportedly owns about 1% of Anthropic, roughly half the stake held by co-founder and CEO Dario Amodei, according to Bloomberg. Schmidt invested tens of millions of dollars in the AI lab's Series A round in 2021, when PitchBook data valued the startup at $461 million. Anthropic is widely expected to go public before the end of the year, with top-end estimates valuing it at as much as $2 trillion. At that valuation, Schmidt's stake could add as much as $20 billion to his net worth, which Bloomberg puts at $57.5 billion. Schmidt and Anthropic did not immediately respond to requests for comment. Read more

OpenAI has acquired Orchestra Bio, a four-year-old AI biotech software startup, the company announced Tuesday on its website. Orchestra builds AI agents that help biotech companies manage drug-discovery operations, from tracking finances to resolving scheduling problems with clinical trials. OpenAI says it is bringing the startup's team and technology into Rosalind, its AI model for life sciences research, and Rosalind Workbench, a tool designed to help researchers plan experiments and analyze results. Terms of the deal were not disclosed. Orchestra had reportedly raised around $20 million from backers including Andreessen Horowitz, Atria, and 8VC. Read more

Boston AI company ASI has raised $325 million in a Series C round that values it at $3 billion, the company announced Tuesday. Spark Capital led the round, with participation from Citadel Securities, Meritech Capital, Craft Ventures, Energy Impact Partners and Andreessen Horowitz. ASI builds software that models air traffic, power grids and military operations hours, days and weeks ahead so operators can act before problems spread. The company recently won an $875 million Federal Aviation Administration contract to manage U.S. air traffic with AI. The new funding will take ASI further into energy, logistics and military field operations, and expand its research and engineering teams in Boston. Read more

Google on Tuesday released TorchTPU, its version of the popular open source toolkit PyTorch, letting developers run AI apps on Google's own chips and cloud. The move follows a similar release by Amazon, which announced its own PyTorch version, TorchNeuron, a day earlier. Both companies are working to loosen Nvidia's hold on the AI industry, where most AI apps are built by default to run on Nvidia's graphics processing units. Industry observers say the shift will not immediately pull developers away from Nvidia hardware, but it gives them easier paths to alternative chips and clouds. Read more

Meta has removed the word “train” from the commercial terms of service that take effect October 30, following a backlash over language that legal experts said could let the company endlessly scrape websites linked from business pages. The company did not alert users to the change. Experts and publisher groups say the revised terms still allow broad use of scraped content to develop AI business products, and news organizations in the U.K., Europe and the U.S. are urging regulators to step in before the policy takes effect. Read more

OpenAI has rolled out a feature called Locked Chats that lets ChatGPT users hide private conversations from their regular chat history, pinned chats and search results. Users can protect a locked chat with a four-digit PIN or, on supported iPhones and Android phones, with Face ID or Touch ID. The company says the feature is meant for moments when someone else may see a screen, such as when handing a phone to a friend or family member. It does not add end-to-end encryption, and OpenAI says it doesn't change how long chats are retained. The feature is rolling out gradually, starting Tuesday, and may take a few days to appear in some accounts. Read more

Anthropic is having trouble securing enough cloud computing capacity from Amazon Web Services, Google and Microsoft, which increasingly see the Claude maker as a competitor rather than a customer, according to an S-1 filing leaked to The Wall Street Journal. The company believes it could spend more than $750 billion on cloud contracts, yet it still worries that newer data center operators may be less reliable. The filing also shows an $80.7 billion net loss for the first half of the year, against $16.3 billion in revenue. Anthropic is reportedly targeting a November IPO. Read more