What is privilege review in eDiscovery?

Privilege review is usually the most sensitive and most expensive part of document review. A responsiveness error produces an irrelevant document; a privilege error can waive protection over confidential legal advice.

What is being looked for. Broadly, two categories: communications made for the purpose of giving or receiving legal advice, and material prepared in anticipation of litigation. The precise formulation and terminology vary by jurisdiction, and in Canada the equivalent concepts are solicitor-client privilege and litigation privilege.

How it is usually done. A first pass narrows the population, traditionally using search terms built from attorney names, law firm domains and phrases associated with legal advice. Documents that hit are then read by lawyers, who confirm the call, apply redactions where only part of a document is privileged, and capture the metadata needed for the privilege log.

Why it is difficult. Term-based screening is blunt in both directions. It over-captures every business thread that copies in-house counsel, and it misses privileged discussion that never names a lawyer. In-house counsel who wear both legal and business hats make this harder still, since the same person's emails may be privileged in one thread and ordinary business correspondence in the next.

Where AI helps. Models that assess the substance of a document can rank the population by likelihood of privilege and surface legal-advice language that term lists miss, then draft log entries for attorney review. The output is a prioritisation signal, not a determination. Privilege remains a legal judgment, so teams validate by sampling the low-scored population, confirming every withheld document, and documenting the workflow.

Related reading: Privilege Review in the Age of Generative AI.

Claira surfaces the specific passage behind each privilege signal so a reviewing lawyer can confirm or overturn it. You can try it on a sample set.