The Toronto-based legal and financial tech company said its revenues rose nine per cent year-over-year in the second quarter, yielding a 28 per cent bump in operating profit compared with the same period in 2025. But shares fell nearly 10 per cent on Wednesday afternoon, after executives faced questions on an earnings call about its ambitions to rely more on Thomson, a custom large language model (LLM). (The Logic)

Talking point: Thomson Reuters is one of many firms facing more questions about how much money they are willing to pour into training their own custom AI models instead of relying on tools from companies like Anthropic and OpenAI, which are increasingly targeting professional sectors. CEO Steve Hasker called the $40-million price tag for training the Thomson LLM “modest,” and a “fraction of the cost” of models from frontier labs, though he said that investment so far has enabled the model to be trained on just 10 per cent of the legal data the company owns. When asked how long it will take for the company to start making money from the custom LLM, Hasker told analysts that the company is optimizing value for “the long term versus any particular quarter.”

Sponsored Content

The road to risk-smart regulation

By Deborah Aarts

Why cybersecurity needs to match pace with AI

By Jessica Aftimus Rosa

How to shift your workforce from AI experimentation to adoption

By Jessica Aftimus Rosa

Paid promotional content