TL;DR:

  • UK law firms have been pouring hundreds of millions into AI tools — with private-equity backing — but the “cracks are starting to show”: Sullivan & Cromwell apologised to a US federal judge this week after submitting AI-hallucinated case citations, and UK courts have repeated warnings for the same class of error.
  • Legal-AI startup valuations have reached extraordinary levels: Harvey at $11bn with Suits actor Gabriel Macht as brand ambassador; Legora at over $5.5bn with Jude Law fronting its “Law just got more attractive” campaign across City AM and London.
  • The combination of AI hallucination risk, client-data sensitivity and AI-enhanced fraud targeting — UK firm Stewarts has already been impersonated in scams — now tops the Law Society’s list of sector challenges.

UK law firms have been pushing hard on AI deployment, and the first reality-check is arriving. City AM’s weekly legal column by Maria Ward-Brennan documents this week’s Sullivan & Cromwell apology to a New York federal judge over AI-hallucinated case citations in a restructuring filing — a striking admission from a firm that charges $3,000 per hour and whose head of restructuring Andrew Dietderich wrote to the court acknowledging the errors.

UK City firms have been “ploughing hundreds of millions of pounds” into AI, with tech spend increasingly underwritten by private-equity funding. The resulting AI-startup boom has produced headline valuations — Harvey at $11 billion, Legora at over $5.5 billion. Both firms have turned to celebrity marketing: Harvey with Gabriel Macht, who played Harvey Specter in Suits, and Legora with actor Jude Law fronting “Law just got more attractive” campaigns that have run across City AM itself.

The Sullivan & Cromwell incident is not isolated. UK courts have had to review cases after lawyers relied on AI-generated citations that proved fictitious, and a senior High Court judge last year warned of cost orders, regulatory referrals and even contempt or criminal proceedings for lawyers making such errors. The emerging discipline inside firms — Sullivan & Cromwell told the court it instructs lawyers to “trust nothing and verify everything” — is the right rule, but requires working hours that erode the productivity case for AI in the first place.

Cybersecurity is the bigger worry

The second reality check is cybersecurity. A Law Society report names cybersecurity as the defining challenge facing UK law firms. Stewarts Law has reported that criminals have been impersonating the firm in emails and faxes — one sent to a City AM staffer was pretending to represent claimants in a fraud matter. Law firms hold large volumes of sensitive client information and handle substantial amounts of money, making them prime targets for fraudsters who are themselves now armed with AI.

The timing is pointed. The Bank of England’s CMORG meeting this week specifically called on UK financial firms to strengthen defences against Anthropic’s Claude Mythos; professional services face precisely the same pressure. NCSC chief Richard Horne’s “warning shots” framing applies to law firms with equal or greater force — the hallucination problem and the fraud problem together now form a material sector-level risk.

Looking forward

UK legal-sector buyers should expect three things over the next six months: tighter AI-use disclosure requirements in professional indemnity insurance wordings; clearer expectations from the Solicitors Regulation Authority and its successor AML supervisor on AI usage documentation (especially relevant given Coadjute’s FCA sandbox entry this week); and continuing downward pressure on Harvey and Legora’s valuations if high-profile hallucination cases continue. Three such cases in the next quarter would change the legal-AI funding conversation materially.

Share this article