Your firm's AI policy problem isn't the tools, it's the missing training

69% of legal professionals now use generative AI for work, more than double the 31% who did a year before. 54% of their firms have given them no training on how to do it responsibly, and have no plan to change that. That's the actual finding from the 8am 2026 Legal Industry Report, a survey of more than 1,300 legal professionals conducted last autumn. Adoption outran governance by a wide margin, and most firms have not closed the gap. They've written a policy, if that, and called it done.

A policy is not training. That distinction is the whole article.

The adoption curve nobody governed

Individual lawyers adopted AI faster than their firms could build any structure around it. Only 46% of firms have rolled out general-purpose AI tools at the organizational level, and just 34% have adopted anything legal-specific, according to the same 8am survey covered by the American Bar Association's Law Practice magazine. That means a large share of the 69% figure is lawyers using ChatGPT, Gemini, or Claude on their own initiative, on personal accounts, with no firm oversight at all.

Only 11% of firms require mandatory AI training. Only 9% have a written policy that's actually enforced, as opposed to one that sits in a shared drive nobody opens. 46% of respondents named data security as a significant adoption barrier, which tells you the concern is there. What's missing is anyone acting on it before a mistake forces the issue.

Why this is a Rule 1.1 problem, not an HR problem

Comment 8 to ABA Model Rule 1.1 requires a lawyer to keep abreast of "the benefits and risks associated with relevant technology." That language has governed technology competence since 2012, originally aimed at e-discovery and cloud storage. It applies to AI with no modification needed. An untrained lawyer feeding client facts into a tool they don't understand, or filing an AI draft without knowing what to check, is not navigating a grey area. They're failing a competence duty that predates generative AI by over a decade.

Roughly 40 states have adopted some version of this technology competence duty. Far fewer have gone further and mandated AI-specific training or CLE credits, which means most lawyers are currently relying on general professional judgment rather than any specific standard for what "competent AI use" requires. That ambiguity cuts against you, not for you, if something goes wrong. Good luck arguing to a disciplinary board that you exercised reasonable competence with a tool you'd never been trained on.

We covered the policy side of this in how to write an AI policy for a small law firm. A policy sets the rules. It does not teach anyone to follow them, and the 8am data shows most firms stopped at the document.

The stakes just went up in Florida, and it won't stay there

On May 28, 2026, the Florida Supreme Court amended Rule of General Practice and Judicial Administration 2.515(d)(2), effective June 15, 2026. Anyone signing a court filing in Florida must now certify that the legal authorities cited in it exist and are accurately quoted. The court didn't ban AI drafting or require disclosure of AI use. It went straight at the actual failure point: unverified output going into a filing under an attorney's signature.

That's a narrower rule than a blanket AI disclosure requirement, and that narrowness is the point. It doesn't matter whether a lawyer used AI. It matters whether they checked what it produced. A trained lawyer treats every AI-drafted citation as unverified until they've opened the case themselves. An untrained one trusts confident-sounding output, and Florida just made that specific habit sanctionable on paper, not just by informal court frustration.

We've already documented what happens when that verification step gets skipped in AI hallucinations in legal work: sanctions, cases, and how to avoid them. Every case in that piece traces back to the same root cause: a lawyer who didn't know, or didn't check, what the model actually got wrong.

CLE content hasn't caught up to how lawyers actually use AI

Even where AI-specific CLE exists, most of it teaches the wrong layer. A large share of approved AI CLE content covers how large language models work conceptually and general ethics and risk awareness, hallucination, privacy, confidentiality, the kind of material a lawyer needs once, not the workflow-level training that changes daily behavior. Agentic tools that chain drafting, research, and review steps together with minimal human prompting are now in active use at major firms, and approved CLE content addressing that specific workflow risk is scarce across the largest providers.

For a solo or small-firm lawyer, the practical result is that CLE credits satisfy a bar requirement without necessarily building the specific skill you need: knowing which of your firm's AI use cases require verification, at what point in the workflow, and what a red flag actually looks like when a model produces it. Treat the credit as a floor. Build the actual habit yourself, or with a colleague who already has it.

What real training looks like for a firm this size

You don't need a training department. You need three specific things, and all three take less than a day to build.

First, a verification checklist tied to each tool your firm actually uses, not a generic "double-check AI" reminder. If a paralegal uses AI for a first-pass contract summary, the checklist says exactly what gets independently confirmed before that summary reaches a client or a filing. Second, a short recorded walkthrough, fifteen minutes is enough, showing a real example of AI getting something confidently wrong in your practice area. Abstract warnings don't stick. Watching a fabricated citation get caught does. Third, a standing rule that any AI output going into a client-facing document or a court filing gets a named human sign-off, logged somewhere, before it goes out.

None of that requires new software or a compliance hire. It requires deciding to spend the afternoon on it instead of leaving the policy document to do work it was never built for.

FAQ

For the underlying rule text and a template to build your own policy from, see our free guide. For a vetted list of tools worth training your team on in the first place, see the best AI tools for lawyers.