Thomson Reuters has begun transferring selected legal AI tasks from Anthropic’s Claude to Thomson-1, its in-house model built through the Snowdon project by adapting Alibaba’s open-weight Qwen. The initial focus is high-volume, structured document review, with legal professionals, accountants and other users of CoCounsel and similar products not expected to see major operational changes. Chief Technology Officer Joel Hron said the move was driven partly by the high cost of systems from Anthropic and OpenAI, while stressing that Thomson-1 will complement rather than replace Claude, which remains central to CoCounsel after Thomson Reuters expanded its Anthropic partnership in May. Thomson Reuters says Thomson-1 performs competitively with Claude Opus 4.8 and ahead of GPT-5.5, Claude Sonnet 5 and Gemini 3.1 Pro across its broader evaluation suite, although its published benchmark results are mixed and independent academic testing is still underway. The project gives Thomson Reuters greater control over its AI stack but raises questions about the security, auditability and political risks of using Chinese open-weight technology. Critics including Anthropic, Senator Tom Cotton and Stanford researcher James Landay have raised concerns about model distillation, security and the limits of treating downloadable model weights as fully open or auditable.