AI Training for Law Firms: A Practical Guide

Solicitors are already using generative AI day to day, often without a firm-wide policy telling them how, and the regulator has taken notice. The Solicitors Regulation Authority’s warning notice on the misuse of AI sets out two risks it expects every firm to manage: AI systems producing false case citations and fabricated legal authorities, and client confidentiality breaches where information entered into a public AI tool effectively places it in the public domain. Training that ignores either risk is training that misses the point of buying it.

What the SRA actually expects

The warning notice does not prescribe a specific training course or a set number of hours. Instead, it takes an outcomes-focused approach: firms and individuals remain accountable for verifying anything AI-assisted before it reaches a client or a court, for maintaining effective human oversight and professional judgement, and for putting contractual and technical safeguards around client data before using any AI system. Using AI does not diminish a solicitor’s personal accountability for the result, whatever tool produced the first draft.

Why one course for the whole firm falls short

A partner, an associate and an in-house counsel are accountable for different things, so a single generic AI session serves none of them well. A partner needs to set the firm’s policy and supervise how juniors actually use these tools day to day. An associate needs the verification habits that keep a piece of work safe before it goes anywhere near a client. An in-house team, with no billable hour to soften the workload, needs to know what genuinely still requires a lawyer and what can safely be templated. Training built around one of these roles does not automatically transfer to another, however good it is for the audience it was written for.

Partners and practice leaders: the supervision gap

Adoption inside firms is often ahead of governance: fee earners pick up AI tools for drafting and research faster than firms write the policies to govern that use. That gap is exactly what the SRA’s warning notice targets, and it is also where training earns its keep fastest. A course built for partners and practice leaders needs to cover setting a usable policy, supervising how associates and paralegals actually use these tools, and answering to clients who ask what the firm’s own safeguards are, not just the theory of what AI can do.

Associates and fee earners: the verification habit

The SRA’s clearest concern, courts criticising lawyers over fabricated case citations, is a verification failure, not a technology failure. Training for fee earners, such as our AI Skills for Legal Professionals course, needs to build the habit of checking every AI-assisted citation and factual claim against a primary source before it is relied on, in the same way a junior would check a precedent before including it in a brief. A course that teaches prompting well but skips verification has taught half the job, and it is the half that carries the least risk.

In-house legal teams: a different set of constraints

In-house counsel work under constraints a firm associate does not: a small, fixed team, high contract volume, and no ability to bill extra hours when the workload increases. Training here is less about the underlying risk, which does not change, and more about triage: what can be handled with a template or a playbook, what still needs a lawyer’s judgement, and how to manage external firms and vendors who are themselves adopting AI at different speeds, sometimes faster than the in-house team itself.

Litigation and disputes: a narrower, higher-stakes use

Disclosure and e-discovery already use AI at real volume, through technology-assisted review of large document sets. The training gap here is specific: knowing where a human reviewer must still look regardless of what the tool surfaces, and having a concrete verification method for anything AI-assisted before it reaches a court, given that this is exactly the use the SRA’s warning notice was written to address.

What good AI training for a firm actually covers

Whatever the role, the SRA’s own framing points to what training needs to include: how to verify AI-generated content against a primary source before relying on it, how to keep client information out of open, public AI tools, and where personal accountability sits regardless of which tool produced the first draft, echoing the same priorities behind the EU AI Act’s AI literacy duty for any UK organisation with an EU footprint. A course that covers prompting technique but not these three has covered the easy half of the subject.

What a phased rollout looks like

Most firms do not need to train everyone on everything at once. A workable order is the foundation course first, for every fee earner regardless of seniority, so the firm has a baseline record it can point to immediately. Role-specific courses for partners, associates and in-house teams follow once that baseline is in place, prioritised by whoever is closest to client-facing risk day to day. Litigation and disputes teams, given the volume of AI-assisted document review already happening there, are usually worth training early rather than last.

A short checklist for choosing a course

Ask whether the course separates partners, associates and in-house teams, or tries to serve all three with one session. Ask whether it addresses the two risks named in the SRA’s own warning notice directly, hallucinated citations and confidentiality, rather than AI risk in the abstract. And ask whether it is dated, since regulatory guidance in this area is still developing and a course written a year ago may already be behind the regulator’s own position.