🔓 Mythos 1: Anthropic dares to release the «too dangerous» model

The model that Anthropic described two months ago as "too dangerous to release" just appeared in their API as claude-mythos-1-preview. Mythos 1 is ready to launch — with controls, but ready. Meanwhile, xAI's finances are a black hole, and one of the most beloved minds in the AI ecosystem just joined the fastest-growing team. The same week, three moves that redefine the board.

🔓 Mythos 1: Anthropic dares to launch the "too dangerous" model

Just hours ago, several users started detecting something strange in Anthropic's API. The claude-mythos-1-preview model briefly appeared visible before being pulled. It's not a bug: Anthropic is preparing the release of Mythos 1 for Claude Code and Claude Security — the model they described in April as "too dangerous to release."

For those who haven't followed this story: in April, Anthropic introduced Claude Mythos Preview, a frontier model with such advanced cybersecurity capabilities that they decided not to make it public. They reserved it for Project Glasswing, their security initiative for critical infrastructure. The community nicknamed it "the model we can't use." But something has changed.

According to leaks, Mythos 1 comes with 1 million context tokens and is being tested internally for deployment in Claude Code (development environment) and Claude Security (defensive cybersecurity). It won't be an open chatbot like regular Claude — it will remain a restricted product. But the fact that it's being prepared for a controlled release suggests Anthropic believes they can now manage the risks. Or that competition is forcing their hand.

And while Anthropic decides to launch its most dangerous model, Musk's company shows that money doesn't always buy talent.

xAI: the ecosystem's black hole

xAI's finances have become public, and the picture isn't pretty. The company is burning cash at an unsustainable rate, its revenue isn't growing as expected, and talent flight is constant. The star hire of the week went to Anthropic, not xAI: one of the most respected researchers in the sector left Musk's company to join Anthropic's security team. The reason, according to close sources: "he wants to work on real security, not headlines."

The contrast with Mythos 1 couldn't be more revealing. Anthropic has the most dangerous model and the most coveted talent. xAI has the noise and the losses.

A model that's scary, a company burning money, and a researcher choosing sides. Three moves in one week that paint an AI divided between those building carefully and those building in haste. History won't say who got there first — it will say who got there intact.

— Max