Do we like them?

Do we like Anthropic?

AI company behind the Claude chatbot and models, based in San Francisco; founded 2021 by ex-OpenAI staff.

Our score: 6 out of 10. A safety-minded AI lab that has stood up to the Pentagon, but it took pirated books, lobbies hard, and is racing ahead on risky tech. We'd back it with open eyes.

Anthropic began in 2021 when a group of OpenAI employees left to build AI with safety as the focus. The name means 'relating to humanity'. Dario Amodei is the CEO. It's now heading for what could be one of the biggest IPOs ever, filed to list on Nasdaq as 'ANT'.

The money is huge and so is the spending. CNBC says it made about $17bn in Q3 revenue and lost $23.29bn that quarter. Yahoo Finance says it has roughly $895bn of compute commitments. Amazon and Google are major backers and suppliers, though their stakes are non-voting and capped.

The big black mark is books. Authors accused Anthropic of using pirated copies to train Claude. Anthropic settled for $1.5bn, which Bloomberg Law calls the largest US copyright class action settlement. A judge gave final approval. Anthropic still says training on books is fair use.

Then the Pentagon fight. Anthropic wouldn't let Claude be used for fully autonomous weapons or mass surveillance of Americans, and the Pentagon labelled it a supply chain risk. A California judge ruled that unlawful retaliation. A D.C. appeals court later upheld the Pentagon's designation, so it's still unresolved.

It's also playing the political game: Forbes says it spent $3.13m on federal lobbying in 2025, and it gave $20m to a group pushing for AI regulation. Add an IPO filing that warns of safety risks, and we land in the middle-to-good: better than most, not a saint.