Arnold & Porter

While TAR is routinely used for responsiveness review, many attorneys still rely on linear, document-by-document human privilege review because traditional TAR has been less effective at identifying privileged documents. In a large antitrust litigation document review (~500,000 documents), that approach would have been costly and slow relative to newer generative AI options.

The firm used Relativity aiR for Privilege, an LLM-based tool, to assist privilege review. Attorneys spent nearly two weeks classifying law firms, attorneys, and third parties as aligned, adverse, or neutral so the model could make privilege predictions. aiR produced overall privilege predictions, privilege categories, and rationales; humans then validated results through targeted sampling and QC, especially for nuanced and borderline categories.

Relative to the estimated cost of a full linear review with contract attorneys, the author achieved approximately 75% cost savings using aiR. The tool’s high-recall first pass helped surface potentially privileged documents and prioritize QC by category, but did not eliminate human review. Cost-effectiveness still depended on dataset size, Relativity’s per-document LLM fees, and how much human QC was required.

• • Generative AI privilege review can deliver large cost savings vs. linear review, but only when upfront entity classification and post-AI human QC are planned into the workflow.
• • Treat aiR as a high-recall first pass: use category-based targeting for QC rather than trusting predictions alone, especially on borderline and ‘in between’ privilege buckets.
• • Before feeding confidential client data into cloud AI, review vendor data-handling terms and ensure supervising attorneys understand the tool’s limits to meet competence and supervisory duties.