Thomson Reuters has postponed the launch of its AI legal assistant, CoCounsel Pro, after discovering the tool was generating fabricated case citations during testing. Originally scheduled for release in early 2025, the company has pushed back the timeline to address these critical accuracy issues. This setback highlights the ongoing challenges tech companies face when developing AI tools for specialized professional fields like law, where precision is paramount.
The AI assistant, built on OpenAI’s GPT-4 technology, was designed to help legal professionals with tasks like contract analysis and legal research. However, during testing, the system produced hallucinations - a term for when AI generates false information presented as fact. Thomson Reuters’ decision to delay rather than proceed with a flawed product demonstrates the high stakes in legal AI, where incorrect information could potentially lead to professional malpractice. The company has not announced a new release date but emphasized its commitment to ensuring the tool meets rigorous accuracy standards before making it available to customers.
Source: https://www.businessinsider.com/thomson-reuters-legal-tech-ai-chatgpt-test-2025-11