Thomson Reuters has a message for the AI world: it would rather own than rent.
The company unveiled its first proprietary AI model, Thomson-1, on Monday. The model is built on Snowdon, a system created by reworking Alibaba's open-source Qwen model.
Open-source software is free to download and change, which is exactly what the company did. The initial focus will be on document review, a task that suits its legal and professional services business.
The move is a direct response to the rising cost of AI services from Anthropic and OpenAI's Codex. Instead of paying those companies every time, Thomson Reuters wants to build its own intellectual property.
CTO Joel Hron put it in simple terms. "Renting a house, you still have a roof over your head, and somebody's taking care of it, and it's great," Hron said. "But you're not building any equity that compounds into something valuable for you long term."
When big firms prefer owning their tools, you can own your future with the Always Be Buying E-Book
Thomson-1 will take over some duties previously handled by Claude, but it will not completely replace Anthropic or any other AI provider just yet.
Here is the twist. In May, Thomson Reuters expanded its partnership with Anthropic for CoCounsel, its legal assistant that currently depends mostly on Claude.
Now the company says it wants to bring more of that work in-house. "Our main objective is to make Thomson the model that powers more and more of CoCounsel's capabilities over time," Hron said.
The irony is hard to miss. Anthropic has claimed Chinese labs are illegally extracting its AI outputs to train their chatbots, and it has pushed for US restrictions. Yet here is a major American company building its own model on Chinese open-source technology.
For investors, this is a story about who controls the tools that power your work. When a company owns its model, it stops paying rent to someone else.
Hron says there is nothing permanent about the tie to Qwen. "There's nothing that necessarily ties us to Qwen," he said.
The team at Imperial College in the UK spent several months adapting Qwen to be "ethically and politically de-biased and safe to use." Not everyone is comfortable with the trend, though. Senator Tom Cotton has warned that US companies using Chinese open-source models could be exposed to security risks like backdoors.
Airbnb, which was named, pushed back. It said it mostly uses American models and that its few Chinese models run through US cloud providers with "privacy and security first" in mind.
For the average investor, the takeaway is simpler. The cost of AI is falling, and the tools are becoming more common. That means more choices, but also more questions about who built the technology and what it can really do. Watching how a company like Thomson Reuters handles that balance could tell you a lot about where your own money is headed.
Instead of renting, learn to build steadily with the free Always Be Buying E-Book