Thomson Reuters is set to launch its own AI language model named 'Thomson,' developed using Alibaba's Qwen technology. According to The Decoder, the company is investing approximately $40 million over a two-year period to build this proprietary model.
The Decoder reported that Thomson's language model achieves top benchmark results primarily when it has access to Thomson Reuters’ exclusive content, such as Westlaw. CTO Joel Hron emphasized that the key is knowing which intelligence to own rather than the intelligence itself, highlighting the strategic value of proprietary data integration.
This move comes amid increasing competition in AI development from players like OpenAI and Anthropic. For Japanese markets, where legal, financial, and data-driven services are rapidly adopting AI, Thomson Reuters’ tailored approach could influence local adoption of specialized AI models in finance and equities.