Law firms are shifting from buying AI to building it, with A&O Shearman paying for its AI strategy largely in billable hours. Legal AI vendors are racing to own proprietary data and in-house workflows, while courts keep sanctioning lawyers for unchecked AI output. Verification and data ownership are the common thread.
Top story: A&O Shearman has committed $1.1bn over five years to AI, mostly as lawyers’ billable time, with every practice giving at least 3% of its hours to building AI tools.
A&O Shearman pays for AI with lawyers’ billable hours
City A.M. · Strategy
The Magic Circle firm will invest $1.1bn over five years, with each practice spending at least 3% of billable hours building specialist AI tools in its own area of law. It is more than double the $500m Kirkland & Ellis pledged in May, and the firm says the billable hour will lose importance as pricing changes. For UK practitioners, the signal is that AI capability is becoming a partner-level, fee-earning priority rather than an IT project.
https://www.cityam.com/ao-shearman-lawyers-ordered-to-sacrifice-billable-hours-to-build-ai-tools/
Legal AI bidding war erupts over collapsed Weil’s data
Law Gazette · Legal
AI-native firms including Crosby, London-based Scissero and Vetta have bid for the tech, IP and legal precedents of Weil. The Gazette calls it the first public bidding war for legal proprietary data. It shows that in legal AI, proprietary data is increasingly the real differentiator, which matters when firms weigh vendor lock-in and what their own precedent banks are worth.
https://www.lawgazette.co.uk/features/legal-ai-what-is-the-end-game/5128467.article
Harvey and Legora both launch agentic in-house intake tools
Law Gazette · Tools
Legora and Harvey launched agentic AI products called Intake for in-house legal teams on 13 and 14 October. The Gazette also notes Microsoft’s Copilot for legal now includes more than 20 legal skills, with plug-ins from leading legal AI vendors. General counsel should expect AI to start triaging business requests directly, and should compare platforms on data and integrations rather than headline features.
https://www.lawgazette.co.uk/features/legal-ai-what-is-the-end-game/5128467.article
AI hallucinations keep landing lawyers in court, with bigger penalties
Virginia Lawyers Weekly · Risk
Two Oregon attorneys were assessed $110,000 in fines and fees for filings full of hallucinated citations. Sullivan & Cromwell also admitted that a bankruptcy filing contained nonexistent case citations and misquoted authority. Penalties are escalating, so supervision and citation-checking protocols are now a core risk-management issue for every firm using generative AI.
EU AI Act high-risk deadlines pushed to December 2027
European Commission (Digital Strategy) · Regulation
The AI Omnibus took effect on 27 July 2026. It delays standalone high-risk AI obligations to 2 December 2027 and embedded-product rules to 2 August 2028. Transparency duties under Article 50 already apply, and the changes still matter for UK firms advising EU-facing clients or deploying AI affecting EU users.
https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force