Thomson Reuters wants to be the AI platform for lawyers. Can it pass the ChatGPT test? (www.businessinsider.com)

Thomson Reuters is leaning hard into generative AI to defend its legal-tech crown, embedding models into flagship products like Westlaw Advantage and CoCounsel to automate research, drafting and agentic workflows. The company reported its legal unit grew organic revenue 9% to $700M and said AI features drove “double‑digit” growth in CoCounsel, even as shares fell more than 6% after the update. Thomson Reuters touts a unique asset: one of the largest legal data troves (ingesting >300 million documents a year) plus editorial human‑in‑the‑loop processing that adds headnotes, fact verification and tags to roughly 85% of primary documents — capabilities it argues are hard to replicate and essential where accuracy, timeliness and security matter. The big question for the AI/ML community is whether Thomson Reuters is building a defensible, model‑level advantage or merely a “model wrapper” that sandwiches proprietary data around general LLMs (it already uses OpenAI’s GPT among others). Competitors — LexisNexis+Harvey, Clio+vLex and the specter of an OpenAI vertical legal app — are pushing the market to benchmark specialized stacks against powerful generalist models like ChatGPT. Technically, the battleground is retrieval-augmented generation, agent orchestration, citation fidelity and human validation pipelines; success will hinge on combining provenance-rich retrieval, robust grounding, and workflows that meet the high‑stakes correctness and auditability lawyers demand.