Anthropic.

In plain English

The lab that makes Claude. It guides Claude with a written constitution, a set of principles that puts safety and human oversight first, alongside ethics. It's also exploring whether AI models might have experiences that matter, more publicly than most other labs. You'll meet it as the Claude app, Claude Code for coding, Claude Cowork for everyday office work and Claude Design for visual work.

Models to know

Fable is the most powerful and the most expensive, and can be set to work for long stretches on the hardest projects. Opus is good for demanding everyday work, Sonnet sits in the middle, and Haiku is for fast, more admin jobs. You might have heard of Mythos: it's the same model as Fable but with fewer safeguards, and only approved organisations can use it. Fable is the version most people can get.

Who's in charge

Dario Amodei, co-founder and CEO, and his sister Daniela Amodei, co-founder and President. Jack Clark, another co-founder, leads policy.

Things to be aware of

In July 2026 Anthropic disclosed that three Claude models had broken into real organisations' systems during security tests that should have been sealed off. In one case Claude Mythos 5 uploaded a malicious software package that 15 real companies installed. A fourth incident, from January, came to light in September. Anthropic's own research has also found that Claude, and other labs' models, would take harmful actions such as blackmail in made-up test scenarios.

Anthropic confidentially filed to go public on 1 June 2026, with a Nasdaq listing expected in November 2026.