Key takeaways
- Thomson Reuters acquired Safe Sign despite zero revenue
- Founder built the startup while training as a lawyer
- Specialized AI research proved valuable without paying customers
Alexander Kardos-Nyheim was still training to become a lawyer when he sold his artificial intelligence startup to Thomson Reuters in August 2024. The company had no paying customers, generated no revenue and had never established a conventional commercial business.
Yet the acquisition became a milestone for Thomson Reuters, marking the first time in its 174-year history that it purchased a pre-revenue company.
The startup, Safe Sign Technologies, had developed specialized AI models for legal work. Its story reveals why some technology companies are acquired before they have sold a single product.
From a Personal Legal Battle to Building an AI Startup
Kardos-Nyheim's interest in law began long before the acquisition. As a teenager, he helped challenge a property development threatening the building where his family lived in southeast London.
The experience exposed him to the difficulties ordinary people face when navigating complex legal disputes without substantial financial resources.
He later studied law at Cambridge and began training at Allen & Overy, now A&O Shearman. In February 2022, he founded Safe Sign Technologies while pursuing his legal career.
Rather than creating another legal chatbot using existing commercial AI models, he assembled researchers from Cambridge, Harvard, MIT and Google's DeepMind to develop language models specifically trained for legal reasoning.
According to Thomson Reuters' acquisition announcement, the startup's models demonstrated promising performance in specialized evaluations.
The approach contrasted with startups focused on raising capital and acquiring users as quickly as possible, including Character.AI, whose later financial difficulties showed that user growth alone cannot guarantee commercial success.
Why Thomson Reuters Bought a Company With No Revenue
By 2024, Thomson Reuters was investing heavily in generative AI for lawyers, accountants and other professionals.
Its CoCounsel assistant already offered AI-powered research and document-analysis capabilities, but the company wanted more direct control over the models underlying those services.
Safe Sign brought something difficult to develop quickly: a team combining legal expertise with advanced machine-learning research.
Thomson Reuters acquired the company on August 21, 2024. Financial terms were not disclosed, although Kardos-Nyheim subsequently indicated that the transaction had transformed the founders' financial circumstances.
The deal illustrates a broader AI talent acquisition strategy, where technology companies pursue researchers and intellectual property alongside traditional revenue-generating businesses.
It also highlights why ownership matters. Companies relying on external models can face licensing restrictions or changing commercial terms, similar to the Microsoft–OpenAI relationship.
The Acquisition Became the Foundation of a New AI Model
After the acquisition, Safe Sign's researchers joined Thomson Reuters' AI development operations.
Their work contributed to Thomson, a proprietary foundation model introduced in 2026 for professional applications including legal and tax research.
Thomson Reuters disclosed that developing the model required approximately $40 million in training and related investment, including talent and computing resources. That amount should not be confused with Safe Sign's undisclosed acquisition price.
The strategy reflects a wider push toward owning specialized technology and protecting valuable intellectual property.
For businesses deploying AI in sensitive professions, reliability and control can matter more than having the largest general-purpose model. As highlighted by the enterprise AI liability debate, companies must consider the consequences of exposing confidential information to external systems.
