Thomson Reuters Unveils Beta of 'Trustworthy' Legal AI, CoCounsel
Thomson Reuters launched CoCounsel Legal AI beta, focusing on reliable, auditable tools for legal workflows.
Why it matters: Legal teams face pressure to adopt AI that is not just capable, but whose outputs can be audited and verified—critical for regulatory, risk, and professional obligations. Thomson Reuters' approach aims to address these requirements, but lawyers say independent validation remains essential.
- CoCounsel Legal AI beta is built using Anthropic’s Claude Agent SDK, a developer toolkit that powers advanced AI assistants.
- 'Fiduciary-grade' outputs are designed to be trustworthy, traceable, and independently auditable within law firm workflows.
- Thomson Reuters claims 1M+ professionals in 107 countries use CoCounsel, but figures haven’t been confirmed by external research.
- Second opinion: Dr. Sara Hastings, AI governance researcher, calls for external audits before legal teams fully trust such systems.
Thomson Reuters has launched the beta version of CoCounsel Legal AI, emphasizing what it describes as "fiduciary-grade" reliability for AI-generated legal work. In this context, 'fiduciary-grade' refers to outputs that can be trusted, traced, and audited—an approach designed for lawyers facing regulatory and professional scrutiny.
- The new platform is built on Anthropic’s Claude Agent SDK—a toolkit that lets developers build sophisticated AI agents able to handle legal queries with detailed reasoning steps.
- CoCounsel natively integrates with the company’s core legal research databases, including Westlaw and Practical Law, so users can cross-check AI work against established resources.
- Thomson Reuters says CoCounsel is intended to act like a highly capable junior associate, but with the added ability to record and audit every step for compliance and risk reporting.
Steve Hasker, Thomson Reuters CEO, said, "CoCounsel is built for moments when being almost right is not good enough... validated by domain experts, and backed by a clear commitment that customer data remains theirs." However, Brooke Conkle, partner at Troutman Pepper who previewed the product, described the experience as “eye-opening” for litigation but noted the need for more peer review and independent evaluation before large-scale legal deployments.
Dr. Sara Hastings, an independent AI governance researcher, echoed this caution: "Any claims of auditability and reliability in legal AI need rigorous third-party assessment before law firms can rely on them in high-stakes contexts." Without external audits, she warns, legal teams should be wary of fully trusting such tools.
Thomson Reuters reports over one million professionals across 107 countries are using CoCounsel, but these adoption numbers have not been independently verified.
By the numbers:
- 1 million+ — Thomson Reuters' reported user base for CoCounsel Legal AI
- 107 — Countries where CoCounsel is in use, per company data
- 2026 — Beta launch year for next-gen CoCounsel Legal AI
Yes, but: AI governance experts and industry partners caution that external audit and peer review will be needed before such tools can meet professional standards for reliability.
What's next: Legal and industry groups are expected to issue independent evaluations of CoCounsel AI as part of adoption reviews later this year.