Regulator receives 42 AI complaints as it warns law firms over oversight

The Solicitors Regulation Authority has received 42 reports of AI-related wrongdoing in the space of a year, and has responded with a warning notice reminding firms that using artificial intelligence in law firms changes nothing about the professional standards expected of them.

The reports were made between July 2025 and July 2026. Several are still being investigated. The issues flagged fall into three broad groups: inaccurate legal citations produced by AI tools, inadequate supervision of the people using them, and breaches of client confidentiality.

What the regulator is worried about

The SRA's first concern is hallucination — legal research, advice, analysis or court submissions that contain cases, quotes or authorities the model has simply invented. Over the past year a string of firms and individual solicitors have been referred to the regulator after false material generated this way ended up in front of a judge.

The second concern is confidentiality. The SRA says it is seeing case details entered into public AI tools such as ChatGPT, which carry none of the safeguards a law firm would expect of a supplier. Its guidance is blunt about the consequence: doing so is likely to breach client confidentiality, and legal professional privilege may be permanently waived and impossible to recover.

Even firms building or trialling in-house systems are not automatically safe. The guidance asks them to satisfy themselves that client data stays inside a secure environment, is not reachable by unauthorised third parties, is not used to train models unless that has been explicitly authorised and is appropriate, and is not kept for longer than it needs to be.

Supervisors are in scope too

One line in the guidance deserves attention from anyone running a team. Where junior or non-authorised colleagues put false citations before a court without adequate review, the SRA says supervisors may themselves be found to have breached regulatory requirements and professional duties.

That reframes AI governance as a management issue rather than a technology one. The question a regulator will ask is not which tool was used, but who checked the output and what the firm's process required them to do.

Aileen Armstrong, the SRA's executive director of strategy and policy, said AI is already benefiting consumers and legal professionals, with firms using it in innovative ways to improve services and efficiency, but that its use does not change professional standards. Firms need appropriate oversight and controls so clients receive high-quality work, she said, and individuals remain responsible for the advice they give whether or not AI was involved.

The profession's response

The Law Society welcomed the guidance, framing AI as something that should support legal practice rather than replace the solicitor's role in giving trusted advice.

Vice president Brett Dixon said that as a regulated profession solicitors have a duty to act in clients' best interests, which includes using AI and other technologies responsibly, and described the warning notice as a reminder of how that applies in practice. He added that innovation in legal service delivery is moving quickly, and the SRA will need to keep issuing swift, clear guidance so solicitors can use AI confidently and in line with their duties.

What law firm leaders should do now

The notice is deliberately not a restriction on using AI. It is a statement that the existing rulebook already covers it. Practical steps that follow from the guidance:

• Write down which AI tools are approved, and ban client data from anything that is not on that list
• Require a named human to verify every citation and authority before it leaves the firm
• Make supervision of AI output an explicit line in supervisors' responsibilities, not an assumption
• Check what your AI supplier does with your data: retention, access and model training
• Record the checks. If a complaint lands, the file is the defence