Two of this month's five free Independent AI Audits are still open, included when you start an AI health check.Check eligibility →
For legal teams

The tool can draft. The lawyer still signs.

Keep client material out of public AI tools, check every citation, and be able to show both.

Research, drafting and review now run through AI, and the courts and the Solicitors Regulation Authority have made clear that accountability stays with the lawyer. GenWA Guard checks AI work against your rules and routes it to the right reviewer, and GenWA Evidence keeps the record for the file.

94%

of UK lawyers say they use AI, while most say they worry about hallucinations (invented answers delivered with confidence).

LexisNexis survey of UK lawyers, 2026 (vendor survey)
18 of 45

authorities cited in a witness statement in a 2025 High Court case were fictitious. The court referred lawyers to their regulators.

Ayinde v Haringey and Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin)
42

reports of possible AI misuse reached the Solicitors Regulation Authority between July 2025 and July 2026.

Solicitors Regulation Authority warning notice, 17 August 2026, as reported by law-firm commentary
March 2026

Insurers have begun asking for an AI policy, training and monitoring evidence at renewal.

Lockton, reported by Legal Futures, March 2026
What you are carrying

Your name is on the work, whichever tool produced it.

18

of 45 cited authorities were invented in one 2025 High Court case, and lawyers were referred to their regulators.

Ayinde v Haringey and Al-Haroun v Qatar National Bank [2025] EWHC 1383 (Admin)
"Someone pasted a client document into a public tool."

The Solicitors Regulation Authority's 2026 warning notice names confidential client information entered into AI tools without safeguards as a core concern. What it costs you: privilege, trust and a possible disciplinary finding.

"A citation we did not check was wrong."

The High Court has said AI tools are not capable of reliable legal research without checking. What it costs you: wasted costs, referral to the regulator and a reputation that is slow to rebuild.

"We cannot show who supervised the draft."

Accountability for AI output sits with the regulated individual and the firm. What it costs you: no record to show a client, insurer or regulator that a named person checked.

"The insurer is asking questions."

Professional indemnity insurers are asking for an AI policy and monitoring evidence at renewal. What it costs you: a harder renewal and a scramble for answers.

"Associates use whatever is fastest."

Without an approved route, people use personal accounts and consumer tools. What it costs you: use you cannot see, on matters you are responsible for.

"Clients want to know our position."

Panel reviews and tenders now ask how AI is used on client work. What it costs you: answers from memory that may not match practice.

A different way to look at it

Banning AI does not remove the risk. It removes your view of it.

A ban moves use to personal accounts, where nobody supervises it, and gives up the time AI can return on research, first drafts and document review. The firm that can say yes with rules it can show is in a stronger position than the firm that says no and hopes. That means an approved route for each task, client material kept inside it, a named checker for anything that leaves the building, and a record the supervising partner can open.

Control applies to AI used through GenWA. Supports your compliance work and is not legal advice.

How GenWA helps legal teams

Six ways to use AI on client work and still show your supervision.

Guard checks the work. Evidence keeps the record. Advisers, human and agentic, shape both to your practice.

From draft to approval

Documents are produced alongside AI use, then flagged to a named person to approve.

When AI helps draft a document, the supporting record is generated with it: what was asked, which rules applied and what was produced. Anything consequential is flagged to the person with authority to sign it off.

Animated flow showing documents produced alongside AI use and flagged to a named person for approval

  1. 1Document generated

    A record is produced alongside the AI use, as the work happens.

  2. 2Flagged

    Anything that needs a decision is sent to the person who owns it.

  3. 3Approved by a person

    A named person approves, changes or rejects it. AI does not sign off.

  4. 4Recorded

    The decision, the approver and the date are kept with the document.

GenWA Guard and Evidence. Steps shown in order.
Use AI when
  • Summarising large document sets for a lawyer to read.
  • Drafting from precedent where a lawyer reviews every line.
  • Sorting and classifying documents at intake.
Keep with the lawyer
  • Advice to a client and the final word on any authority cited.
  • Decisions on credit, hiring or medical matters.
  • Anything that must be identical every time, such as a deadline calculation.
A tool can find the authority. The lawyer confirms it exists.

The line is not set by the technology. It is set by who is accountable and how quickly a mistake would be caught.

Where it matters most

Three settings where legal teams feel it first.

Pick the setting closest to yours.

Privilege, citations and supervision

The Solicitors Regulation Authority expects accountability to stay with the firm. Guard keeps client documents inside approved routes and sends citations to a named checker, and Evidence records who checked what for the matter file.

See legal

Client files and professional indemnity questions

Practices handle sensitive financial data and answer to insurers and professional bodies. Managed accounts, Rules and the Log show how AI-assisted work is supervised.

See accountancy

Contracts, consent and records in a regulated setting

In-house counsel in care providers advise on special-category data, consent and incident records. Rules and the Data-Flow Map show where health data goes, and the Log supports an inspection response.

See healthcare and care
Ways of working with us

Three routes for a legal team, from a first review to a supervised build.

Advisers are human and agentic. A fourth route, the Accelerator, is by application only and is paid by equity, project, retainer or licence, for founder-led businesses that want us alongside them.

Review load calculator

Estimate how much checking Guard could take off your team.

Enter how many documents your people check each month and how long a check takes today. This is an indicative estimate from your own figures, not a forecast.

Review workload calculator

Drafts, replies, summaries, decisions support: anything AI produces for use.
Set this by your own rule, for example anything customer-facing or consequential.
Used only to show the workload as a share of one person's time.

Workload estimate, from the numbers you enter

Reviews required each week40
Review hours each week3.3 hours
As a share of one person's working week8.9%Workload estimate
Assumptions
  • Every output that needs review is reviewed once, by one person.
  • Who must review, and what counts as consequential, is your decision, not an estimate.

Items marked assumption can be edited. No figures are stored or sent.

How this is worked out

Reviews a week = outputs a week x share needing review. Review hours = reviews x minutes per review / 60. Share of one person's week = review hours / working hours in a week.

Bring them with you

Three colleagues who share the risk, and three functions to bring alongside.

The people who carry accountability, and the teams working beside legal.

Want to see how a supervised draft looks?

Book a demo and we will walk through a document, its check, its approval and the record, using example data.

Questions we are asked

Straight answers.

Can law firms use AI safely?

Yes, when AI use is limited to approved routes, client material is protected and a named lawyer checks the output. The Solicitors Regulation Authority has said it does not tell firms to avoid AI, but that accountability stays with the solicitor and the firm.

How do we stop client documents going into public AI tools?

Give people an approved route that is easier than the public tool, and apply rules before a request leaves. GenWA Data Protection applies identity, rules and redaction to AI used through GenWA, so personal data is redacted before approved external model calls.

How can we catch invented citations?

Make verification a named step before anything is filed or sent. Guard can flag documents that contain citations and route them to a named checker, and Evidence records who checked them. The checker is a person.

What will our insurer want to see?

Insurers are increasingly asking for an AI policy, training and monitoring evidence. Evidence Packs gather the register, rules and exceptions for a period into one report you can hand over.

Does GenWA give legal advice?

No. GenWA supports your own compliance work with controls, checks and records. Legal advice stays with lawyers.

Does it work for a firm of ten to fifty people?

Yes. A firm of ten people or more can start with a short review, one approved route and a simple record, without a large team. Advisers, human and agentic, shape it to the size of the practice.

Where is the record kept?

The record is held in your AWS or Azure account, or hosted by GenWA on AWS as agreed. We agree which in writing before we start.