Thomson Reuters buys pre-revenue AI firm Safe Sign for millions

Thomson Reuters has acquired Safe Sign Technologies, an artificial intelligence research startup, for millions. Closed just twenty months after the company was founded, the deal represents the 174-year-old information provider’s first pre-revenue acquisition.

The transaction stands out in a market obsessed with rapid commercialisation. Thomson Reuters’ venture arm moved within minutes of reviewing Safe Sign’s internal results, validating a strategy that deliberately ignored revenue in favour of building a proprietary AI model focused on safety, robustness and reliability.

That model was developed on a shoestring budget by a small team drawn from Cambridge, MIT and Harvard. The founder, who trained as a solicitor at Allen & Overy while building the company on the side, initially tried to build a consumer legal product. When that failed to attract funding, the startup pivoted entirely to deep research.

Securing capital required looking overseas. The founder flew to New York with £200 after UK investors repeatedly passed on the business. Nearly all of Safe Sign’s eventual funding came from North America. Reflecting on the disparity, the founder noted that American investors typically asked "how can I help?" whereas British ones asked, "how will this fail?".

Now an active investor, the founder argues the market consistently mistakes style for value. "We are trained to admire the loud things: revenue climbing on a chart, capital raised, a founder on a stage. Institutions, it turns out, reward none of that. They reward you for doing the work so thoroughly that, when someone finally looks closely, there’s nothing to find but substance."

This thesis points to a distinct funding gap in the AI sector. While capital floods into massive frontier models and famous labs, the founder argues the real opportunity lies in the underlying infrastructure these labs depend on.

Specifically, this includes testing infrastructure to ensure models are not simply gaming evaluations, solving the fundamental machine learning problem of continual memory, and building specialised tools for early AI-for-science labs. "Style calls these tooling, storage, and services. Substance says they’re three of the harder problems in the field... And almost none of the capital swarming into AI is going in this direction."

For market participants, the Safe Sign exit serves as a case study in mispriced technical risk. It suggests that pre-revenue startups solving complex, unglamorous problems for the AI industry may offer more defensibility than those chasing immediate top-line growth.