Thomson Reuters Launches Own AI Model, Raises Revenue Outlook

Revenue Beat and AI Gamble: Thomson Reuters Bets on Its Own Model

Thomson Reuters lifted its full-year revenue growth projection to roughly 8% after second-quarter sales rose 9% to $1.95 billion, slightly ahead of analyst estimates. The stronger outlook was accompanied by a strategic push into artificial intelligence that goes far beyond licensing someone else’s model. The company disclosed that it has developed a proprietary large language model called Thomson, already benchmarked against systems from Anthropic, OpenAI and Google, and tailored specifically for legal research, tax analysis, accounting and audit tasks.

Chief Executive Steve Hasker described the opportunity to create a “sovereign AI” solution for the firm’s largest and most demanding customers – one that could also strengthen flagship products like Westlaw and Co-Counsel. The model is designed to handle the exacting professional work that those platforms already support, while giving clients the data control they frequently demand when entrusting sensitive materials to AI. Hasker framed the move as a way to accelerate service delivery, scale more efficiently and build a cost advantage over time.

The AI push follows a recent decision to sell a majority stake in the company’s news-wire division, refocusing the entire group on digital services and AI-driven professional tools. By bringing model development in-house, Thomson Reuters signals it is no longer content to be a mere reseller of generative AI – it intends to embed the technology into the core of its long-standing information and workflow businesses.

Why Thomson Reuters Built Its Own AI — and How It Reshuffles the Legal Tech Race

Data Sovereignty as the Decisive Motivator

The CEO’s emphasis on “sovereign AI” points to a real customer anxiety. Law firms, accounting networks and corporate legal departments routinely handle privileged, confidential and personally identifiable data. Many have been hesitant to pipe that information into third-party AI services run by startups or even big-tech labs, fearing loss of control. Thomson Reuters can offer the model inside its own trusted, audited infrastructure – the same one that already hosts Westlaw and its compliance suite. That alone could differentiate its offering from a generic GPT wrapper.

Reshaping the Legal AI Competitive Landscape

By benchmarking Thomson against Google’s and OpenAI’s models, the company is telling the market that it can match general-purpose AI performance while adding domain-specific tuning. For law firms, the choice is no longer between generic tools and nothing; it’s between a dedicated legal-AI platform integrated with their existing research workflows and a bolt-on from a different provider. This puts direct pressure on legal-tech start-ups that have built their own AI research assistants, as well as on rival information providers such as LexisNexis, which are also investing in generative AI. Thomson Reuters’ installed base of Westlaw and Practical Law users gives it a distribution advantage that a standalone AI tool cannot easily replicate.

The Economics Behind Building Rather Than Buying

Developing a proprietary model is expensive, but for a company with $1.95 billion in quarterly revenue and margins buttressed by subscription information services, the calculus is appealing. Over the medium term, owning the model can lower the per-query cost compared with paying API fees to an external AI vendor, while also allowing tighter integration and faster iteration. The sale of the Reuters News division freed up both capital and management attention, giving the AI strategy the focus it needs at the top of the organization.

What the Thomson AI Push Means for Law Firms, Accountants and Rivals

• For law firms and tax practices: Evaluate whether the Thomson model’s native integration with Westlaw and Co-Counsel beats your current workflow, and ask for concrete data on how it handles jurisdiction-specific legal reasoning, not just broad benchmarks.
• For corporate legal and accounting departments: Scrutinise the data-governance promises. If the model runs on the same infrastructure as your existing Thomson Reuters subscription, it may simplify your internal AI-usage policies and ease concerns about data leakage.
• For investors: Watch for the first client deployments and any indication of ARPU uplift from AI features, especially when the company next reports. A successful AI upsell could move the revenue growth trajectory above the just-raised 8% forecast.
• For competitors: The sovereign AI argument is a direct challenge to any rival selling a cloud-native legal assistant. Expect differentiated data-residency and ethical AI narratives to become table stakes, not add-ons.