Overview
The AI sanctions wave has reached the top of the profession. On July 1, Vice Chancellor Lori Will of Delaware’s Court of Chancery ordered Richards, Layton & Finger, one of Delaware’s most prominent firms, to show cause why it and one of its directors should not be sanctioned. Days later, a British legal taskforce warned of liability running the other way, against lawyers who decline to use AI at all. And in-house legal departments, which now use AI more than any other corporate function, expect the lawyers they hire to keep up. That leaves the profession a needle to thread, with sanctions for careless use on one side and negligence exposure for refusing AI on the other. The space between is narrow, but the court just mapped it.
At issue in Delaware is a familiar story with a new ending. A brief in Leiske v. Kidd, revised with generative AI, went out the door with “fictitious citations, fabricated quotations, and hallucinated legal propositions.” The duty to verify, Vice Chancellor Will wrote, is “nondelegable,” and counsel “cannot discharge those obligations by relying on a paralegal, much less on an artificial intelligence program.” Her order requires the firm to account, under oath, for how it governs AI.
The British warning came six days later, from the UK Jurisdiction Taskforce. Its new legal statement on liability for AI harms concludes that a professional can be negligent for declining to use AI where a reasonable professional of comparable rank and specialism would have used it. The statement’s example speaks directly to litigators. A solicitor might breach the duty of care by failing to advise a client to consider AI-assisted review of a large document set.
The statement addresses England and Wales, but American lawyers already carry a version of the duty. Forty states have adopted a duty to keep abreast of “the benefits and risks associated with relevant technology.” The UKJT statement shows where that duty may lead.
Clients may get there first. Legal and governance departments are the heaviest AI users in the enterprise, logging 19.5% of AI hours in Harmonic Security’s usage index. Yet when Deloitte surveyed 121 senior in-house legal leaders this spring, 58% said their outside providers rarely or never raise AI’s benefits unprompted, and only 4% had experienced those benefits themselves. In an Association of Corporate Counsel survey of 657 in-house counsel, 59% report no AI savings from their firms, and 61% plan to press for changes in pricing and delivery. When the heaviest users of the tools are the clients reading the bills, they know exactly what is possible, and they get frustrated watching outside counsel leave it unused. That frustration means pressure to adopt swiftly, albeit carefully.
The Chancery order itself maps the safe path. Using generative AI in court filings, it explains, is “not inherently problematic,” provided the lawyer “carefully verifies the accuracy of the output.” And the three demands it put to Richards Layton, now answered by affidavit, double as a checklist for everyone else: the firm’s “written policies, guidelines, and restrictions” governing AI, a record of how those policies were “communicated,” and the “technological safeguards” and “training programs” that keep future filings accurate. The time to have those answers is before a court asks.
The needle can be threaded. Use tools you can trust and verify their outputs. Measure twice, cut once, and don’t stab yourself with the needle.
The Weekly Inference is a recurring Step Into IP feature. Nothing here is legal advice.
