Thomson Reuters unveiled its own AI model, Thomson-1, on August 24, 2026, aiming to reduce its reliance on the expensive Anthropic AI systems it previously depended on for legal document review and consulting tasks [1]. The Thomson-1 model adapts Alibaba’s open-source Qwen large language model and integrates Thomson Reuters’ proprietary legal knowledge base to provide specialized legal consulting tools [1]. Initially, Thomson-1 will handle document review functions formerly carried out by Anthropic’s Claude model, though Reuters has no current plans to fully replace Anthropic products [1].
Anthropic, meanwhile, is preparing to launch a closely watched initial public offering within weeks, expected to be among the largest ever on Wall Street [2, 3]. The AI company, which focuses heavily on tools for software programmers such as its Claude Code assistant, reported annualized revenue of about $65 billion in July 2026, up from $47 billion in May but below some investor expectations of $80 billion [2, 3].
Consistent with national security scrutiny, the Trump administration in March 2026 suspended government contracts with Anthropic, citing supply chain risks, a move that remains tied up in legal disputes [2, 3]. Anthropic’s flagship AI model Fable 5, launched in early June 2026, has struggled to gain traction with enterprise customers, accounting for just 11% of company AI tool expenditures two months after launch [1, 3]. High pricing and U.S. government restrictions have limited access, dampening demand [3].
Anthropic currently serves around 6,000 enterprise customers, each spending over $100,000 annually on its AI services [3]. OpenAI’s July 2026 launch of GPT 5.6 at lower pricing has helped OpenAI grow its annualized revenue by 35% in the third quarter of 2026, intensifying competitive pressure on Anthropic [3]. The market also now offers numerous cheaper open-source models from China and other regions, increasing competitive challenges for companies reliant on proprietary technology [1, 3].
Miles Clements, a partner at Accel, noted that "most people don’t really need cutting-edge technology. The era of blind demand chasing the most advanced models will quickly end" [3].
Anthropic is expected to file its IPO documents in the coming days, marking a key milestone as Thomson Reuters bet on a hybrid AI approach combining open-source models with expert domain knowledge for its new Thomson-1 platform [1, 2, 3].
