Anthropic is a US AI lab whose Claude models underpin European deployments tracked here, including Allianz's claims-automation platform. In July 2026 it hired Monzo cofounder Tom Blomfield to run compute strategy, betting that compute access, not model architecture, is the binding constraint on more autonomous AI. The hire landed weeks after a 19-day access blackout for European customers, the dependency on US-controlled infrastructure that Mistral and Germany's SPRIND programme are trying to counter.
In September 2026 it published a machine-checked formalisation of Fermat's Last Theorem in Lean, 13 million lines produced by parallel Claude agents in eleven days, built on Mathlib and Kevin Buzzard's multi-year project at Imperial College London. Days later Matt Clifford resigned as chair of Britain's ARIA, five days after joining Anthropic as managing director of international affairs, after Parliament rejected the conflict-of-interest safeguards the UK business department had approved.
On 12 September Dario Amodei said Anthropic would give third-party evaluators permanent, employee-level access to its systems, and proposed a coordinated industry slowdown of one to two years. Two days later the company launched Claude for Financial Advisors, wired into BlackRock, Charles Schwab, Vanguard, and iCapital, saying the model would not give investment advice; under MiFID II a personal recommendation can be implied rather than stated, and ESMA has told firms that using AI changes none of their suitability obligations.