Law firms should make AI training mandatory, not optional
Pinsent Masons’ safeguards, a wave of firm-side AI promotion, and recent sanctions all point to the same fix: supervised training has to become part of ordinary professional duty.
An AI-assisted editorial, reviewed by a human before publishing. It reasons over our own tracker data (and cited context) — a point of view, not legal advice.
The strongest argument against my view is easy to state: lawyers already know how to check their work, and the recent AI sanctions show what happens when they do not. Fair enough. But that is exactly why firms should stop treating AI training as a perk, a pilot, or a glossy page on the website. The problem is not mystical. It is process. Pinsent Masons said this week that it has put safeguards around its internal AI pilot after the court incident involving AI-generated errors. That is the right instinct, but it should not be exceptional. It should be the floor.
This week’s firm pages read like a confession if you look at them together. Pinsent Masons says it piloted Lexis+ AI for legal research, drafting, and review, while measuring time savings and user adoption. Gowling WLG has an AI, Innovation and Knowledge Team, and its director of practice technology leads a team that supports AI-enabled legal service delivery and strategic investment decisions. Icertis says its Dioptra acquisition is the next step in its AI strategy, with pre-signature agents for playbook creation and first-line review of new agreements. The industry is building capability in public. The question is whether it is building judgment at the same pace.
The next sanctions wave will come from poor process, not bad intent.
The market is training people faster than it is training responsibility
That imbalance is not a theoretical worry. Bloomberg Law reported that Big Law firms are pushing associates to use AI tools in creative ways, including competitions, consultant advice on an AI mindset, and even giving billable credit for learning how to use generative AI in daily work. At Ropes & Gray, two first-year associates each month present to partners on how they have been using AI, and those presentations are part of a program that lets them devote 20% of their billable hours to learning generative AI. That is serious institutional effort. It is also, in its own way, an admission that firms know training matters more than slogans.
Law.com went further and put the real risk plainly: from Australia to the U.S., firms are pouring money into AI while hiring fewer entry-level lawyers, and the bigger risk is failing to develop the judgment, strategy, and client skills future partners will need. That is the part firms keep dodging. If you thin out the apprenticeship and then flood the shop floor with tools that can produce plausible nonsense, you do not get efficiency. You get a pipeline of lawyers who know the software and have never been required to prove they can supervise it.
The sanctions wave is a process problem, not a novelty problem
The disciplinary cases already show what this looks like when it goes wrong. A junior solicitor submitted false citations generated by an AI tool in Google Scholar, and the work was not checked. Steven Schwartz and Peter LoDuca filed a ChatGPT-produced brief with fictitious cases. Jan Tomasik and Daniel Mann filed bogus ChatGPT citations and were removed, referred to the Nevada bar, and fined. A California attorney filed ChatGPT-generated fake case quotes and citations and was fined $10,000. None of that required malice. It required bad process and too much trust in a tool that can sound certain while being wrong.
That is why the old firm habit of letting AI training sit in the voluntary bucket is indefensible. If a firm is already choosing to market AI, measure time savings, name a technology team, or build workflow tools around contracts and research, then it also has a duty to teach supervised use in a way that can be checked. Otherwise the firm is doing the easy part in public and leaving the hard part to chance. The next sanctions wave will not come from lawyers trying to commit fraud. It will come from lawyers who were never taught how to review AI output as a matter of routine professional hygiene.
Make the training mandatory, and make it auditable
This is where the profession should get blunt. Every lawyer who touches client work should be required to take supervised, auditable AI instruction, not because AI is magical but because it is ordinary now. Pinsent Masons’ safeguards after its own pilot show that firms already understand internal controls matter. The wider market, from Ropes & Gray’s associate programming to Gowling WLG’s internal team to Icertis’s contract workflow push, shows the tools are moving into real practice. If firms can track time savings and adoption, they can track completion, supervision, and review standards too.
The alternative is a familiar legal self-deception: we will assume the junior knows what to do because the junior clicked the tool. That is how you get fake citations, unchecked drafts, and a disciplinary record that looked avoidable in hindsight. The best AI program in a law firm is not the flashiest one. It is the one that teaches lawyers when the machine is useful, when it is dangerous, and who has to sign off before a document leaves the building. Until firms make that compulsory, they are not training lawyers. They are training excuses.