TL;DR:

• Freshfields has agreed a deal with Anthropic to co-develop specialist legal AI products — for contract drafting, document review and corporate due diligence — that Anthropic intends to sell to rival law firms.
• The magic-circle firm will roll out Claude across its offices globally and gain early access to future Anthropic models, though not the Claude Mythos cyber-capability model whose rollout is still restricted.
• Anthropic’s associate general counsel Mark Pike said the tie-up is “a signal to every other law firm that moving slowly with AI deployment is not an option right now”.

Freshfields has agreed to help Anthropic build legal AI tools that can eventually be sold to other law firms, marking Anthropic’s first partnership of this kind and a visible acceleration of AI adoption inside the top tier of UK and global law.

What the deal covers

Under the agreement, Anthropic’s lawyers will work with Freshfields’ attorneys to develop tools for drafting documents, reviewing contracts and running due diligence on corporate transactions. Freshfields will roll out Claude across its offices worldwide, pay Anthropic an undisclosed sum, and gain early access to future Anthropic models — with the explicit exception of Claude Mythos, whose cyber-vulnerability capability has prompted a deliberately slow rollout. Anthropic cannot use Freshfields’ data to train its models, and Pike declined to say who would own the IP in any jointly-developed tools or whether Freshfields would share in future revenue.

The deal is Anthropic’s first law-firm partnership, and follows similar industry-specific deployments with Accenture and Goldman Sachs. For Freshfields, the appeal is speed: managing partner Alan Mason said the firm wanted to combine its legal expertise with Anthropic’s technical capacity to “come up with better products for ourselves and our clients”. It is also a signal to clients, who increasingly expect in-house lawyers to “do more with less” — a pressure that trickles upstream to the external firms they hire.

The risk lawyers cannot ignore

The timing is pointed. Sullivan & Cromwell, a US elite firm, apologised to a federal judge this week after an important filing contained multiple AI-generated “hallucinations”, including a misquote of the US bankruptcy code. Mason said Freshfields had a process where “nothing goes out without being checked by individual lawyers”, but the wider legal sector is still calibrating what that human review actually looks like at volume. Freshfields separately rolled out Google’s AI tools last year and is building bespoke AI agents for legal work on that stack too, so Claude will sit alongside competing tooling rather than replace it.

Looking Forward

For the UK legal market, the Freshfields-Anthropic deal reframes the competitive question. Magic-circle firms are no longer just early AI adopters — they are now co-developers whose process expertise shapes what commercial legal AI looks like. Expect mid-market firms to face a procurement choice between Harvey (which already has Allen & Overy and A&O Shearman deployments), Anthropic-via-Freshfields derivatives, and Microsoft-backed tools like Edinburgh-based Wordsmith. Billable-hour economics will be the pressure that resolves it.