Legal AI

Westlaw AI-Assisted Research

High risk

Generative AI research feature inside Westlaw Precision that answers legal questions in prose with citations drawn from the Westlaw database.

Verified 2026-08-31Thomson Reuterslegal.thomsonreuters.com

Is Westlaw AI-Assisted Research safe for confidential data?

The data handling is enterprise-grade (same Thomson Reuters stack as CoCounsel: no training on customer content, SOC 2 and ISO 27001, DPA available), so the high rating is about output reliability, which for a research tool is the product. The peer-reviewed Stanford RegLab study (Journal of Empirical Legal Studies, 2025) found Westlaw's AI-Assisted Research hallucinated on roughly a third of tested queries — about twice the rate of Lexis+ AI — while producing complete, accurate answers only around 42% of the time, despite Thomson Reuters marketing that leaned on hallucination-avoidance. Because answers arrive in confident prose with real-looking citations inside the platform lawyers trust most, unverified reliance is the specific, documented failure mode: Damien Charlotin's database of court decisions involving AI-hallucinated material passed 1,600 cases in mid-2026. Use it as a drafting accelerant, never as authority.

Risk by plan

The same product often carries very different terms depending on the tier — consumer plans are where the exposure concentrates.

Westlaw Precision with AI-Assisted Research (only tier)
No training

Available only as part of a Westlaw Precision subscription under Thomson Reuters' standard enterprise terms; no consumer tier.

Data handling

Training on inputs

Thomson Reuters states customer queries and content are not used to train its models or third-party LLMs; LLM providers are contractually barred from training on customer data.

Retention

Third-party LLM calls follow Thomson Reuters' zero-retention pattern; research history is retained in the customer's Westlaw account under subscription terms.

Residency

Processing is U.S.-based; non-U.S. firms should confirm cross-border transfer terms in the Thomson Reuters DPA.

Compliance

  • SOC 2Yes
  • GDPR / DPAYes
  • HIPAA BAANot verified

Certifications typically apply to specific tiers and contracts — confirm scope in writing before relying on them.

New to these frameworks? See our plain-language guides to SOC 2 and the other AI compliance standards.

Enterprise controls

  • SSO / SAML
  • Contractual no-training terms with LLM providers
  • SOC 2 Type II and ISO 27001 certifications
  • Standard Thomson Reuters DPA available

Frequently asked questions

Does Westlaw AI-Assisted Research hallucinate?

Yes, at a documented rate. The Stanford RegLab study, peer-reviewed and published in the Journal of Empirical Legal Studies in 2025, measured hallucinations on roughly 33% of queries — about double the rate of Lexis+ AI in the same test — and found complete, accurate answers only about 42% of the time. Thomson Reuters disputed aspects of the study's methodology, but no vendor has published contrary independent benchmarks.

Is Westlaw AI-Assisted Research reliable enough for filings?

Only as a starting point. Every proposition and citation needs independent verification in the underlying sources before it reaches a filing — ABA Formal Opinion 512 makes clear the lawyer remains responsible for AI-assisted work product, and courts have sanctioned lawyers in a growing body of cases tracked in Charlotin's public database. The tool's citations link to real Westlaw documents, which makes verification fast; skipping it is the documented route to sanctions.

Our associates already use ChatGPT for research — is Westlaw AI at least safer than that?

For confidentiality, considerably: queries stay under an enterprise agreement with no-training terms, instead of a consumer chatbot's defaults. For accuracy, the gap is smaller than the price difference implies, which is why the sensible posture is both a sanctioned, grounded research tool and firm-level controls over what gets pasted into consumer AI — one does not substitute for the other.

Policy changelog

  • Initial entry published from Thomson Reuters' published security documentation and cited independent research.

Westlaw AI-Assisted Research is probably already in your organization.

Sanitized AI shows you who is using it and redacts sensitive data from prompts before it leaves your control.

Get a demo

More AI tool profiles