Why Thomson Reuters is betting on its own LLM to win the legal AI race

Thomson Reuters is betting that its legally-trained LLM will give it an edge in the legal AI race.

We sat down with CTO Joel Hron and AI research co-lead Alexander Kardos-Nyheim to get into why the data giant is doubling down on its own legal-specific LLM, “Thomson”, which will soon begin powering parts of its flagship legal AI product, CoCounsel Legal.

Thomson traces back to the company’s 2024 acquisition of Safe Sign Technologies, the AI research company founded by Kardos-Nyheim while he was a trainee at A&O Shearman. That team now leads foundational legal AI research inside Thomson Reuters.

Hron describes Thomson Reuters as an “expertise company” and says the move builds on the company’s chief strength: combining proprietary data with in-house expertise to improve performance on legal tasks.

Thomson is not replacing models from the AI giants like OpenAI or Anthropic. Instead, it is used alongside them in workflows where accuracy and legal reasoning matter most, such as document review.

Kardos-Nyheim says general-purpose models lack deep training in the law. Thomson, by contrast, is built on datasets including Westlaw and Practical Law, alongside input from Thomson Reuters’ vast network of legal editors. He likens the difference to a liberal arts graduate (who knows a lot about a lot) versus a law firm partner with an entire career’s worth of legal experience.

The first Thomson-powered features are expected to roll out in CoCounsel later this year.