Before using an AI service on a client matter, a UK law firm can make a documented, matter-specific decision about the tool, information and intended task. This helps the firm identify unresolved handling questions, assign responsibility and show what was checked before sensitive material is used. It is an operational control, not a determination that privilege applies.
Check the specific service and configuration
Record the service, account tier, model provider, enabled connectors and relevant settings. Review the terms that apply to the firm's account and check how prompts, files and outputs are handled, including retention, model improvement, subprocessors, access and deletion. Keep a dated record of the terms and configuration reviewed; a product label or general vendor description is not a substitute for checking the applicable terms.
Apply client and matter restrictions
Check the engagement terms, client instructions, outside counsel guidelines and the firm's own information-handling requirements. Identify the information involved and the proposed use. Set out what is permitted, what needs further approval and what must not be entered while the handling or permission remains unclear. Ask the responsible partner or qualified counsel to resolve legal questions.
Assign review and keep an audit trail
Name the lawyer or matter owner responsible for authorising the use. Define how outputs will be checked against primary sources and matter facts, who may access prompts and outputs, and where the decision and review evidence will be retained. The lawyer remains responsible for the firm's work; AI output should not be treated as a substitute for professional review.
Use a repeatable readiness record
For each proposed use, capture the task, information category, approved service and configuration, relevant terms, client restrictions, decision owner, review steps and record location. Revisit the decision if terms, settings, features, matter sensitivity or incident information changes. If a material control cannot be verified, pause sensitive inputs until the responsible owner decides what is safe to proceed.
When should a firm seek an independent readiness review?
If teams are using AI inconsistently, or the firm cannot readily show which tools and controls have been checked, Pattrn Data's AI Risk & Efficiency Audit can help identify operational risks and prioritise practical controls. For implementation support, see Pattrn Data implementation services.
Source and scope
This practical guidance is informed by the source question and the referenced discussion, “When AI Isn’t Privileged: Confirmed SDNY’s Written Opinion Elaborates on Confidentiality, Work Product and Waiver,” published by Subject to Inquiry. The discussion concerns a US legal context; this article does not present it as UK legal authority or offer a conclusion on privilege. Read the source discussion.