Discovery, Privilege, and AI: What Could Go Wrong

Discovery in litigation involves combing through mountains of documents. AI can help speed the process, but if not used carefully, it can expose privileged material, leak confidential data, or undermine case strategy.

How AI is used in discovery

Firms and corporate legal teams are experimenting with AI tools to sort documents, flag relevance, and even suggest privilege calls. This can save time and reduce costs compared to manual review.

Privilege at risk

Attorney–client privilege protects confidential communications between lawyers and clients. But if privileged documents are mishandled by AI tools — especially if uploaded to public or vendor-controlled platforms — privilege may be considered waived. Courts often treat disclosure, even accidental, as loss of protection.

Training and data retention issues

Some AI tools keep prompts and responses for system improvement. If discovery documents are uploaded, that data might remain stored or even influence future outputs. This creates long-term risks if the same vendor serves opposing counsel.

Audit trails and access

Discovery requires strong records of who accessed what, when, and how. Many AI systems lack detailed audit logs, which can make it difficult to defend privilege claims later.

Practical safeguards

Use vetted, enterprise-grade tools: Consumer-grade AI chat apps are not designed for litigation workflows.
Check vendor contracts: Ensure they include data handling, confidentiality, and audit commitments.
Keep sensitive sets internal: Consider on-premise or private deployments for high-stakes matters.
Validate privilege calls manually: AI can flag, but lawyers must make the final call.

Takeaway

AI may change how discovery is handled, but privilege law hasn’t changed. If privileged documents are exposed, the responsibility rests with the legal team. Careful tool choice and human review remain essential.