Thomson Reuters Builds Custom AI Model for Professional Services

How Does the Model Compare to Industry Leaders?

Thomson Reuters has developed its own artificial intelligence model tailored for legal, tax and compliance professionals, trained on the company’s extensive proprietary content library. The model is designed to enhance features within products like CoCounsel and support workflows across its professional information platforms. Internal testing indicates the model performs competitively against leading general-purpose AI systems from OpenAI, Anthropic and Google on various professional and broad capability benchmarks.

The initiative reflects Thomson Reuters’ strategy to reduce reliance on external AI providers while maintaining high accuracy and domain-specific relevance. By leveraging its own curated data—including case law, regulatory documents and expert analysis—the company aims to deliver more reliable outputs for high-stakes professional use. The model integrates directly into existing tools, allowing seamless adoption without disrupting current user experiences. Early results suggest strong performance in tasks such as legal research, contract analysis and regulatory monitoring.

Why Keep Using Anthropic’s Models Too?

Internal evaluations show Thomson Reuters’ AI model holds its own against prominent competitors in specialized professional assessments. While not claiming superiority across all metrics, the model demonstrates particular strength in tasks requiring deep understanding of legal and financial frameworks. This positions it as a viable alternative for organizations prioritizing data provenance and compliance with professional standards. The company emphasizes that benchmarking focuses on real-world applicability rather than generic language tasks.

Despite developing its in-house solution, Thomson Reuters continues to use Anthropic’s AI models in certain applications. This hybrid approach allows the company to balance control over data and model behavior with access to cutting-edge capabilities from external leaders. Specific use cases may benefit from Anthropic’s strengths in The dual strategy supports flexibility and risk mitigation in a rapidly evolving AI landscape.

What is the primary purpose of Thomson Reuters’ new AI model? The model is designed to power AI-driven features in professional products like CoCounsel, focusing on legal, tax and compliance workflows using the company’s proprietary content.

Frequently Asked Questions

Does Thomson Reuters plan to replace all external AI models with its own? No, the company intends to use a combination of its proprietary model and selected external models, including those from Anthropic, depending on the specific requirements of each application.

How was the model trained, and what data was used? The model was trained exclusively on Thomson Reuters’ internal professional content, such as legal databases, tax regulations and compliance materials, ensuring high relevance and accuracy for professional users.