Can using AI tools waive legal professional privilege in the UK?
Putting privileged material into an AI tool does not automatically waive privilege. It can create confidentiality and disclosure risks, so check the tool's terms, access and retention settings, and your firm's matter-specific restrictions before using it.
For a proposed use, identify who receives the material and what the service does with it. Review any legal commentary against the actual matter and processing arrangement; this article does not determine the privilege consequences of a particular disclosure.
Assess disclosure before using privileged material
Assess the proposed disclosure before putting privileged material into an AI system. Check provider and connected-service access, retention and reuse against the matter's restrictions. The legal effect of a disclosure needs a separate assessment; neither a tool label nor this checklist establishes that privilege is preserved or waived.
- Inputs to keep out pending assessment: client advice, draft pleadings and sensitive communications where the tool's terms, retention or reuse have not been approved for that matter.
- Terms and controls to assess: contractual protections, tenant isolation, training use and access controls for the actual product tier and enabled features. An enterprise label does not clear privileged inputs.
Practical rule: keep privileged material out of unapproved AI tools. An NDA or security assurance alone does not settle the disclosure's legal effect.
Book an AI Risk & Efficiency Audit
Where privilege gets exposed: common AI workflows that create waiver risk
Identify which material the firm's legal reviewers consider privileged, including questions of legal advice or litigation privilege. Then map how each proposed workflow exposes that material. Check the following input, account and output paths before approval:
1) Copy-paste inputs that include client identifiers or legal advice
- Draft advice notes, memos, counsel instructions, internal emails that summarise advice.
- Attachments like contracts with tracked changes and negotiation positions.
- Chronologies, witness summaries, interview notes.
2) “Shadow AI” and personal accounts
Include personal accounts, browser extensions and free tiers in the tool inventory. Approval of one workflow does not cover a different account, feature or provider; assess the arrangement actually used.
3) Tool retention, logging, and training defaults
If the provider retains prompts/outputs, uses them to improve models, or allows broad internal access, you may have an avoidable confidentiality problem even if nobody intended disclosure.
4) Sharing outputs externally without checking provenance
Before sending an AI-assisted client email or court document, check the information used to generate it and what the output reveals. Review the intended recipient and matter restrictions. The presence of privileged facts in a prompt does not let this article determine whether a particular disclosure crossed a legal boundary.
Controls to assess for privileged material
Assess the controls below against the matter, staff practices and processing arrangement. They are options for a firm to evaluate, not proof that privilege will be preserved or that a particular setup is approved.
Platform controls to evaluate
- Approved AI stack only: publish a short allowlist of tools and block the rest on managed devices where feasible.
- Contract and configuration: verify any no-training promise, retention limits, tenant isolation and access controls against the actual terms and enabled features.
- Data loss prevention (DLP): test which uploads it can detect or block, including client identifiers, matter numbers and document classifications, and record the gaps.
- SSO and role-based access: access by practice group and matter sensitivity, not “anyone with a link”.
- Logging: decide which records are needed to explain use and review. Avoid unnecessary copies of privileged material; set access, storage and retention rules before collecting prompts or outputs.
Policy guardrails (clear rules that survive pressure)
- Default rule: no client confidential or privileged content in public or non-approved AI tools.
- Explain restricted content: use examples such as advice, drafts, strategy, counsel communications, litigation preparation and internal legal analysis. Have legal reviewers define the matter-specific restrictions; the examples are not a determination that every item is privileged.
- Permitted use cases: general research on public facts, generic drafting, style improvements, checklists, non-client-specific templates.
- Escalation path: if a team wants to use AI on privileged material, require an approval process (risk review plus tool configuration review).
- Client instructions: where relevant, incorporate client consent/limitations into engagement terms or matter plans.
Process guardrails (how work gets done day to day)
- Redaction workflow: review names, dates, locations, unique facts, deal values and matter numbers. Check indirect identification and confidential information remaining after redaction before deciding whether the prompt may be used.
- Matter-based controls: classify matters by sensitivity and restrict AI features accordingly (for example, no external processing for high sensitivity matters).
- Human-in-the-loop review: AI output is never sent to a client or filed without qualified review and a provenance check.
- Prompt templates: provide generic or hypothetical examples that reveal no client information, and explain when a prompt needs matter-specific approval.
- Scenario-based training: show fee earners prohibited inputs and how to request an assessment for a proposed alternative. Use synthetic training examples rather than unnecessary client material.
Operating model: who owns privilege risk for AI?
Name a lead for the firm's privilege-related AI decisions and identify who provides legal, information security and data protection review. The allocation below is a proposal to adapt to the firm's actual responsibilities.
- Practice leadership: defines acceptable use by work type and matter sensitivity.
- Risk and compliance: owns policy, exceptions, incident response, and client requirements.
- IT and security: owns tool approvals, configurations, DLP, access controls, and monitoring.
- Knowledge management: provides templates, safe prompt libraries, and curated internal sources.
If ongoing help is needed, agree who maintains the playbooks, checks controls and reviews evidence after tool or matter changes. Compare an external service with the firm's available internal capacity before choosing an arrangement.
Explore managed AI governance and support
Implementation checklist for the firm's assessment
- Tool decision: pick your approved AI tool(s) and ban everything else by default.
- Contract and settings: confirm no training, retention controls, tenant isolation, and admin visibility.
- Policy: write for fee earners, with permitted and prohibited examples, matter restrictions and an approval route. Choose the length needed to make those rules usable.
- Controls: SSO, RBAC, device management, and DLP for common exfil routes.
- Training: use synthetic scenarios relevant to each practice group and check whether staff can apply the rules. Choose the session length around that need.
- Exception process: a simple form and turnaround time for higher-risk matters.
- Evidence: keep proportionate approval and review records, with access and retention rules. Confirm which records a client or auditor needs; logs alone do not prove privilege is protected.
Reminder: this is general information, not legal advice. The right controls depend on your practice mix, clients, and the AI architecture you deploy.
Get help implementing AI guardrails
Need a safer AI route for your firm?
If this article reflects a live decision in your practice, the useful next step is to map the workflows, confidentiality risks, supplier controls and governance gaps before tools spread informally.