FAQ: Common TAR & AI Questions

Short answers to questions users often ask before using TAR or AI-assisted review.

Question	Practical answer
Is TAR court-approved?	Yes, courts have approved TAR in appropriate cases, but approval depends on a reasonable, proportionate, well-documented workflow.
Do I have to disclose seed sets?	Not always. Transparency obligations vary by court, protocol, and dispute posture. Many teams disclose process and metrics rather than all training documents.
Can TAR replace attorney review?	TAR can prioritize and reduce review burden, but attorney judgment is still needed for coding guidance, privilege, QC, and production decisions.
Is generative AI the same as TAR?	No. TAR usually refers to classification/prioritization workflows; generative AI may summarize, reason over, or draft from documents. They can complement each other.
What makes a TAR process defensible?	Clear scope, consistent coding, documented training, validation sampling, QC, proportionality rationale, and the ability to explain decisions.
When should we stop reviewing?	When validation, proportionality, issue coverage, and case needs support that further review is unlikely to materially change the result.
Can AI review privileged material?	Only with strong controls. Consider confidentiality, vendor terms, data retention, privilege waiver risk, model training, and human review safeguards.