Legal risk meets datacentre build
- Freshfields' deal with Anthropic raised questions about who is liable when legal AI tools produce erroneous advice. - Anthropic is hiring a six‑figure negotiator to secure datacentre deals as it accelerates European expansion. - The combination of rapid infrastructure build‑out and unresolved liability norms heightens legal and operational risk for enterprise AI adopters. ( )
Anthropic is expanding in Europe on two fronts at once: selling AI into legal work and hiring a London dealmaker to lock down datacentre capacity. (cnbc.com) (legaltechnology.com) On April 23, Freshfields and Anthropic announced a multi-year collaboration to deploy Claude across the law firm globally and co-build legal workflows. Freshfields said 5,700 employees already have access through its internal platform, and usage rose about 500% in the first six weeks. (legaltechnology.com) The same day, CNBC reported that Anthropic is hiring a London-based “Transaction Principal” to negotiate European compute-capacity deals. The role pays £225,000 to £270,000 and covers sourcing, developer outreach and term-sheet negotiations across Frankfurt, London, Amsterdam, Paris, Dublin, the Nordics and Southern Europe. (cnbc.com) A legal AI tool is software that drafts, summarizes or answers questions from case files and contracts, like a junior assistant working at machine speed. When a law firm builds that into client work, the central question is whether responsibility stays with the lawyer using it, shifts to the software provider, or is split by contract and negligence law. (lawyer-monthly.com) (eur-lex.europa.eu) In Europe, that question now sits alongside a new rulebook. The European Commission says obligations for providers of general-purpose AI models under the European Union AI Act started applying on August 2, 2025, adding documentation, transparency and risk-management duties for model providers. (digital-strategy.ec.europa.eu 1) (digital-strategy.ec.europa.eu 2) Anthropic’s infrastructure push is moving just as fast as its enterprise sales. CNBC reported this week that the company committed to spend more than $100 billion on Amazon Web Services technology over 10 years and, earlier in April, expanded a Broadcom-linked arrangement for about 3.5 gigawatts of computing capacity using Google processors. (cnbc.com 1) (cnbc.com 2) The company is also building out its physical presence in London. On April 16, Anthropic said it had secured office space in the city for 800 people, up from more than 200 staff already based there. (cnbc.com) European capacity is getting tighter as model builders and cloud groups chase the same power, land and permits. CNBC said Microsoft recently added compute capacity at an Nscale site in Norway, Nebius outlined a large AI factory in Finland in March, and Microsoft and Oracle have also announced datacentre plans in Portugal, Spain and Italy since the start of 2025. (cnbc.com) Public resistance is rising at the same time. CNBC reported on April 15 that at least $156 billion in datacentre projects were canceled or delayed in 2025, citing Data Center Watch, as opposition to AI infrastructure grew in the United States. (cnbc.com) Freshfields says its rollout will follow security, compliance and training frameworks, and Anthropic told CNBC it declined to comment on the European datacentre hiring plan. For companies buying AI tools, the immediate picture is concrete: more powerful systems, more servers behind them, and more legal paperwork around who pays when the output is wrong. (legaltechnology.com) (cnbc.com)
