AI news, 10 November: Anthropic launches Claude Fable 5.2 and Mythos 5.2 for ambitious technical work

Anthropic releases Claude Fable 5.2 and Mythos 5.2. Anthropic has rolled out two upgraded models aimed at complex coding, scientific research and long-running projects. According to the company, Fable 5.2 is available to all users while Mythos 5.2 (with fewer safeguards on cyber and biology topics) is limited to verified organisations. Anthropic says the models run up to 30% faster than their predecessors, cost less ($8 per million input tokens), and in testing helped coordinate more than 1,000 AI agents to build a working operating system over eight days, formalise Fermat’s Last Theorem in about 30 hours, and tackle open problems in black-hole physics. The company positions them for ambitious technical work rather than everyday chat. Availability includes major cloud platforms.

Anthropic IPO details briefly circulated then retracted. Reuters reported that pricing information for a potential blockbuster IPO—aiming to raise up to about $69.3 billion—was sent to bankers and then pulled back as inaccurate, with an updated filing expected. The episode underscores intense investor interest and volatility around the AI leader’s planned public listing, expected soon, even as the company continues major product pushes.

Meta expands AI age detection worldwide on Instagram. Meta said it is rolling out its AI-powered age-estimation technology globally to identify users it believes are 13–17 and automatically place them in more restrictive Teen Accounts, even if they registered with an adult birth date. Protections include private defaults and limits on contacts and content. The company is extending the approach on Facebook and plans a similar global push for Threads later; parents in the US will receive explanatory notifications. The move follows earlier regional rollouts and legal pressure over youth safety.

Nvidia tightens overseas AI chip cloud access for China. Reports including from Tom’s Hardware and earlier Financial Times coverage say Nvidia has instructed cloud partners to reserve at least 51% of services for non-Chinese clients on new contracts, limiting a key rental route Chinese firms have used to access advanced GPUs. Black-market prices for restricted high-end racks have reportedly soared in China. The step adds to US export controls and compliance pressure amid the US-China AI competition.

Trump signals openness to AI “guardrails.” In a 60 Minutes interview, President Trump said AI risks “can be managed,” named the Justice Department as overseer, and stated he “doesn’t mind guardrails” provided they do not leave the United States in “a deep second position” to China. He also claimed an unnamed major AI executive had been warned of jail. Separately, Nvidia’s board probe into CEO Jensen Huang’s alleged ties to a suspected Chinese intelligence asset continues, with Trump saying the FBI is looking into the matter; the company has said it has no reason to believe secrets were compromised.