AI speeds investigations but raises privilege and evidence risks

Generative AI can accelerate document review, identify suspicious communication patterns and help prepare interview plans and investigation reports. In an Old Bailey murder trial, AI-assisted review of 10,000 documents compressed review time by one month and saved £50,000. These gains require human oversight to protect the accuracy and completeness of findings.

UK regulators are adopting the technology unevenly. The Financial Conduct Authority is rolling out generative tools for authorisations and supervision, while the Competition and Markets Authority uses AI to help detect bid rigging and potential consumer protection infringements. The Serious Fraud Office has trialled technology-assisted review, but detailed public guidance on generative tools in enforcement investigations remains limited.

Legal privilege depends heavily on confidentiality, contractual protections and how tools are used. Secure systems used by lawyers may preserve privilege where other legal requirements are met. Public tools create greater uncertainty, with differing UK and US approaches. Under English law, seeking legal advice directly from a chatbot does not attract legal advice privilege.

Hallucinated citations, altered evidence and fabricated complaints create further risks. Professionals can face negligence exposure for unsuitable tools or inadequate verification and, potentially, for failing to use AI where competent practice requires it. Key safeguards include secure systems, senior legal supervision, independent checks, documented use and prompt correction of errors. Regulatory adoption does not necessarily mean authorities will accept AI-generated submissions.