Compliance Pods for Legal Professionals
All Episodes

AI Misuse in Law: Hallucinations, Privilege, and SRA Risk

This episode breaks down the surge in AI misuse reports to the Solicitors Regulation Authority and the growing fallout from unchecked hallucinated citations in court filings. It also examines how pasting client material into open AI tools can breach confidentiality and permanently waive legal professional privilege.

Show Notes


Chapter 1

The 42 Misuse Reports and the Hallucination Trap

Paul Crowther 2

Welcome back to Compliance Pods for Legal Professionals, and just a reminder that the content of this podcast is for general information purposes only and does not constitute legal advice. So Buckle up. Because: Here. We. Go.

Andre Grayson 2

So, forty two separate formal reports of AI misuse landed on the Solicitors Regulation Authority's desk between July 2025 and July 2026. Forty two in just twelve months, Paul. And these were senior judicial referrals straight from the High Court, plus self reports from major international firms like Pinsent Masons after a junior associate used AI and drafted misleading court letters.

Paul Crowther 2

Pinsent Masons, wow. If a firm of that scale is catching junior associates pasting unchecked AI outputs into court submissions, you know it is happening everywhere. What did the High Court actually say in that Haringey case?

Andre Grayson 2

Right, so in the application of Ayinde v Haringey London Borough Council, the court brought together two cases under its Hamid jurisdiction. And paragraph 29 specifically states that when a lawyer places false citations before the court, whether because AI was used without proper checks or otherwise, it is likely to be appropriate for the court to make a reference to the regulator.

Paul Crowther 2

Under Hamid jurisdiction! That is, uh, that is serious business. That is where judges call senior partner supervisors onto the carpet to explain why the court was lied to.

Andre Grayson 2

Exactly. And the SRA's Warning Notice makes it crystal clear that relying on AI output is not a suitable defence. They cited Brett v The Solicitors Regulation Authority from 2014, reminding everyone that misleading the court is one of the most serious offences an advocate or litigator can commit.

Paul Crowther

Let us talk about why this is actually happening on the ground. You have got junior fee earners under immense pressure to hit six or seven billing hours a day. They see a tool like ChatGPT that can draft a summary in thirty seconds, and they, they, they take the shortcut. But paragraph 3.5 of the Code of Conduct means supervising partners are directly on the hook for those unverified outputs, right?

Andre Grayson 2

Precisely, paragraph 3.5 requires effective supervision. You cannot just blame the trainee or the prompt when fabricated case law gets submitted to a judge.

Paul Crowther 2

And it gets even worse when you look at how people are feeding client data into these systems. Is it true that if you use the paid version of ChatGPT, your data remains private?

Andre Grayson 2

That is a lethal misconception. In UK v Secretary of State for the Home Department, 2026, the Upper Tribunal addressed this directly. At paragraph 21, the Tribunal observed that to put client letters and decision letters from the Home Office into an open source AI tool, such as ChatGPT, is to place this information on the internet in the public domain.

Paul Crowther 2

In the public domain! So by pasting a confidential Home Office letter into a web browser AI, you are literally broadcasting it to the world?

Andre Grayson 2

That is what the Tribunal warned. Using AI tools in this way will likely breach client confidentiality, and as a result, legal professional privilege may be permanently waived and unable to be recovered. Permanently waived, Paul. You cannot undo that paste. Both free to use and paid for systems pose huge risks.

Paul Crowther 2

So unless you have a secure environment where data is not stored, not accessed by third parties, and not used to train models, you are walking through a regulatory minefield.

Andre Grayson 2

The SRA's non negotiable position is that human in the loop verification is mandatory. The lawyer remains personally responsible for every single word in their legal submissions, regardless of the tools used in drafting.

Paul Crowther 2

So when a disciplinary tribunal asks why false precedents or leaked client secrets ended up in a filing, saying the machine hallucinated will just get your practicing certificate suspended that much faster.

Andre Grayson 2

Without a doubt. Verify everything, secure your environment, or face the SRA.

Paul Crowther 2

Alright, good warning for everyone today. So, from all at Legal Compliance Support:, stay compliant, Stay safe and we’ll catch you next time!