What does the SRA's AI misuse warning notice require of a firm?
A warning notice is the SRA signalling enforcement. Read alongside the updated supervision guidance, it expects AI outputs reviewed by a person and supervision the firm can prove.
Answered in short
5 things that decide this
- 01The SRA published a formal warning notice on AI misuse on 17 August 2026, after receiving 42 reports of AI misuse between July 2025 and July 2026. A warning notice signals enforcement.
- 02It pairs with the SRA's updated Effective Supervision guidance of 12 June 2026, which requires human review of AI outputs before they're relied on.
- 03Together the two mark the SRA's move from guidance to enforcement on AI in 2026. The working rule is the one regulators keep repeating: AI may assist, but the lawyer remains accountable for every output.
- 04What a firm must be able to show is supervision in fact: where AI assisted, who reviewed the output, and a record a supervisor could hand the SRA on request.
- 05None of it bans AI. A firm using AI with a review checkpoint and a log sits inside the notice's expectations. A firm using it quietly with neither is the reason the notice exists.
Guidance became enforcement in 2026
The SRA's position hardened in 2026. Misuse reports gave it the caseload: 42 filed between July 2025 and July 2026. Its Effective Supervision guidance update of 12 June 2026 set the expectation that a person reviews AI outputs before anyone relies on them. Then the August warning notice put firms on formal notice: AI misuse is an enforcement matter, not a learning curve.
For a firm, the practical reading is short. Any AI-assisted drafting, research or client work now sits inside your supervision arrangements, and "we told people to be careful" is not an arrangement. Their question is whether review happened, and whether you can show it.
Supervision you can show, in practice
Supervision you can show means three things exist for AI-assisted work. A log records where AI assisted on a matter. A required human check happens before any output leaves the firm. And a supervising partner can hand both to the SRA on request. The checkpoint satisfies the guidance; the log makes your compliance real rather than asserted.
Disclosure positions belong in the same system. Which client is told what differs by matter and by rule. A template mapped to the current guidance beats working out an answer each time.
We build legal AI with visibility as a structural feature. LexPair, a legal lead-generation platform Hashlogics built, matches cases with AI-assisted logic while admins keep full visibility into every lead. A supervision layer for a firm runs on the same principle: the model works, and people see everything it did.
- AI assistsA draft, research or a summary on a matter.
- Solicitor reviewsHuman review before reliance, per the guidance.
- Client positionDisclosure handled per the applicable rule.
- LoggedMatter, tool, reviewer, date.
- On requestThe record a supervisor hands the SRA.
A checkpoint satisfies the guidance; a log proves it ran.
Related questions
01Does the warning notice ban using AI?+
No. It targets misuse: AI-assisted work reaching clients or courts without the review and supervision the rules already require. A firm with a review checkpoint and an audit trail is using AI inside the SRA's expectations.
02Who is accountable when AI-assisted work goes wrong?+
The solicitor doing the work, and the firm through its supervision arrangements. Human review of AI outputs is now a supervisory expectation under the updated guidance. A missing review therefore reads as a firm failure as well as an individual one.
03How does this compare with US bar guidance?+
Same principle, different force. As of June 2026, 15 US jurisdictions have formal AI ethics opinions and 38 have guidance. ABA Formal Opinion 512 is the national reference, and enforcement varies by state. The SRA moved earlier on enforcement: one national regulator, a formal warning notice, and a year of misuse reports behind it.
Related
- Software for law firms →Our legal hub: what firms actually buy.
- AI agent development →Our service page: legal AI built with the checkpoint and the log in place.
- What does ABA Formal Opinion 512 require firms to disclose? →The US ethics-opinion layer, mapped.
- What happens when a lawyer cites an AI-hallucinated case? →The misuse the SRA's reports are made of.
