If you are a doctor in the EU or the UK and you opened the OpenEvidence app this week, you got an error message instead of an answer.

OpenEvidence, the Miami-based clinical decision support tool that 40 percent of US physicians use daily, withdrew its product from the European Union and the United Kingdom in the past few days. The company’s public reasoning, reported through HIStalk and confirmed by user reports across Reddit and Threads, points directly at regulatory uncertainty including the EU AI Act.

It is the first US AI unicorn to publicly pull a live product out of Europe over AI law. It will not be the last.

What OpenEvidence does

OpenEvidence is what AI was supposed to look like in healthcare before the hype cycle ate the conversation. Founded in 2021 by Daniel Nadler (also the founder of Kensho, which IHS Markit bought for $700 million in 2018), the product is a search and reasoning tool for doctors. You ask it a clinical question. It returns answers grounded in NEJM, JAMA, Cochrane, and other peer-reviewed sources, with citations.

It is not a chatbot. It is not a wellness app. It is a working medical reference that has, in three years, become daily-active infrastructure for over a million American physicians.

Forbes reported in January that the company closed a $250 million round at a $12 billion valuation. Nadler’s personal wealth doubled to roughly $7.6 billion in the same week.

And as of this week, none of that is available to a doctor in Berlin, Madrid, Stockholm, or London.

Why they walked

The EU AI Act’s August 2, 2026 deadline turns medical decision support into a high-risk system under Annex III. That triggers conformity assessment requirements, CE marking, technical documentation packages, post-market monitoring obligations, and exposure to fines up to 35 million euros or 7 percent of global revenue.

OpenEvidence operates on a freemium model in the United States. The product is free to verified physicians, monetized through pharma. Putting that business model through a 35-million-euro fine envelope did not pencil. So they pulled out.

The UK is not part of the EU AI Act, but the British government has signaled its own AI bill is coming. OpenEvidence appears to have decided that running two compliance regimes for two markets that account for a fraction of their revenue was not worth the legal team.

What this looks like to a doctor in Munich

Brutal.

An American GP can pull up an answer on Lyme disease prophylaxis or rare drug interactions in 15 seconds. Their German counterpart, as of this week, cannot. The competing tools available in Europe are either national language editions of older medical references, hospital intranets running on PubMed, or generic ChatGPT with no clinical grounding and no liability cover.

The reaction on European medical Twitter and Reddit was not subtle. "It hasn’t been taken down by the EU. It’s been taken down by OpenEvidence themselves, alleging regulatory uncertainty about the AI Act," one Reddit user posted. "This is not helpful," wrote a Mushtaq Bilal post on Threads that has since gone viral inside European academic medicine circles.

The political optics are: a tool that demonstrably saves doctors’ time and probably saves lives is no longer available, and the named reason is European law.

Why the timing is brutal

OpenEvidence pulled out the same week the AI Act simplification package collapsed in Brussels. We covered the trilogue failure in a separate piece this morning. The two stories are the same story.

The Commission’s Digital Omnibus was specifically supposed to soften the AI Act enough that medical, fintech, and HR products could keep operating in the EU without absorbing punitive compliance costs. Wednesday’s no-deal trilogue means there is no softer version coming before August. So a product team at a US company looking at the August cliff has to make a binary call: in or out.

OpenEvidence called "out." Other AI medical players (Hippocratic AI, Glass Health, Abridge, Suki) will be in product-strategy meetings this week running the same math. Some will follow.

The bigger problem

Europe positioned the AI Act as the model for global AI governance. Brazil, India, the UK, and over a dozen US states wrote bills that mirror its risk-tier structure. The argument was that strict rules would build public trust and produce better AI products.

What is happening instead is that strict rules are producing fewer AI products available to Europeans. The first concrete data point of that thesis was OpenEvidence this week.

If three more US AI unicorns follow OpenEvidence out of Europe before August, the political conversation shifts. Right now the AI Act is sold as protecting Europeans from bad AI. After enough geofencing announcements, it starts looking like denying Europeans access to good AI. The latter is much harder to defend in a continental election cycle.

Meta’s Llama, OpenAI’s GPT, Anthropic’s Claude, and Google’s Gemini are still available in the EU because the hyperscalers can absorb compliance costs the way a $12 billion startup cannot. The result, ironically, is that the AI Act may end up entrenching exactly the Big Tech dominance European policymakers spent the last decade trying to break.

What to watch

Three near-term signals.

One: whether other US AI medical players match OpenEvidence and pull out before August. Glass Health and Hippocratic AI are the obvious watch list.

Two: whether any European member state’s health ministry publicly comments on the loss of access. If France, Germany, or the Nordics pressure Brussels for a medical carve-out, the omnibus restarts.

Three: whether OpenEvidence files a formal grievance with the European Commission. They almost certainly have the legal team to do it. So far they have chosen the cleaner path of just leaving.

Daniel Nadler built a $12 billion company on the premise that giving doctors fast, grounded, citation-backed AI would save lives. As of this week, half a billion Europeans are no longer part of that experiment. The reason is not technology. The reason is law.