The AI audit is coming for your paper trail

The profession has spent two years arguing about whether AI is clever. Courts have answered with a more lawyerly question: can you stand behind the paper? On one side, the numbers say this is no sideshow — 30 AI-related sanctions tracked, with New York leading at 5, California at 4, and Oregon at 3. On the other, the cautionary tales are painfully concrete. A defense brief in Coomer v. Lindell contained nearly 30 defective citations and cases that do not exist after unverified AI use. In New York City, a lawyer filed a brief riddled with fake cases after using Microsoft Copilot and ended up fined and referred for discipline. In California, counsel in a custody fight over a dog was fined $5,000 after filing briefs with fabricated cases. This is not an abstract ethics seminar. It is a filing problem.

The machine can draft. The lawyer still signs.

The new liability is embarrassment with a docket number

If you think the penalty for sloppy AI use is only a scolding, the record says otherwise. Justice Joel M. Cohen awarded fees and referred Michael Fourte to ethics committees after briefs with AI-hallucinated citations and quotations in Ader v. JS Property Holdings LLC. A federal judge in Oregon declined formal sanctions against Buchalter after a filing contained fake case citations from AI use, which is hardly a victory lap; it is a reminder that not every bad result gets a formal hammer, but every bad filing still leaves a stain. The common thread is not that AI is banned. It is that the court expects the lawyer, not the machine, to verify what gets submitted. That used to be basic professional competence. Apparently it still is, which is awkward for everyone involved.

Meanwhile, the market is training the nervous system of the profession

The market is not waiting for the bar to catch up. Latham & Watkins says it launched an AI Academy to train lawyers, while also pointing to AI-related deal work on Broadcom’s AI XPV platform and CoreWeave’s $3.1 billion delayed-draw term loan facility. Harvey says Microsoft’s Corporate, External, and Legal Affairs organization — about 2,000 lawyers and compliance professionals — will use its platform for legal and compliance operations. Thomson Reuters’ CoCounsel Legal and Harvey both score 90 out of 100 in the catalog, alongside RelativityOne. Those are not toy scores for toy problems. They are signals that large organizations are treating AI as part of legal infrastructure, not a novelty demo. The point for a working lawyer is not to chase the shiny object. It is to recognize that clients and counterparties are getting comfortable with AI-assisted legal work while still expecting humans to own the result.

The sensible response is not fear, but a paperwork revolution

Firms that understand the moment are not merely buying software; they are committing to process. Kirkland & Ellis has earmarked $500 million of its revenues to develop its own AI platform. Cleary Gottlieb bought Springbok AI and brought AI engineers in-house. Tarter Krinsky & Drogin launched a firmwide Office of AI and Innovation to embed responsible AI, technology, and process innovation across operations. Capital Group is building a global AI enterprise program within Legal & Compliance to create policies, governance frameworks, tools, and scalable systems. That is the real lesson for practicing attorneys: AI is not replacing legal judgment, but it is forcing every serious shop to prove that judgment exists in the first place. The lawyers who survive this shift will not be the ones who prompt fastest. They will be the ones who can document, verify, and explain what happened between draft and filing. The machine can draft. The lawyer still signs.