Europe’s AI firms call the US slowdown pitch self-serving

Europe's AI companies have publicly broken ranks with the US frontier labs' call to slow down, and they are not being polite about it. Mistral, the region's leading AI developer, accuses US incumbents of using the safety moment to "consolidate their market position", while Proton's chief operating officer dismisses the whole exercise as "totally self-serving". The target is Dario Amodei's proposal that leading developers pace their capability gains and seek an antitrust waiver so the biggest labs can jointly develop safety measures under permanent embedded evaluators. To the labs already at the frontier, that reads as prudence. To the companies still climbing, it reads as a rope ladder being pulled up.

Watch Desk analysis

What happened

Amodei's essay urged advanced developers to slow the pace of capability improvements, arguing that legislation moves too slowly to keep up, and proposed that companies voluntarily coordinate on safety standards instead. Mistral's response, in a statement, was that "some incumbents are using this moment to consolidate their market position, pushing for regulation designed to favour them over competitors", adding that the risks in question had been clear for months.

The backdrop explains the heat. Europe has produced few frontier developers, and three-year-old Mistral is widely seen as the region's best hope of challenging US and Chinese models. Most of the continent runs on American systems it would rather not depend on, which makes any incumbent-written rulebook look suspicious before it is even read.

Why it matters

For a fortnight the slowdown debate looked like a closed shop, with Amodei, Sam Altman and Elon Musk all warning about powerful systems and agreeing the pace should ease. Europe's reaction breaks that consensus and changes the question from whether AI should slow down to who benefits when it does.

The concrete flashpoint is the waiver. An exemption from competition law for the biggest labs, with those labs defining what counts as dangerous and supplying the inspectors, is exactly the arrangement a challenger would design if they wanted the race to stay precisely as it is. For a region that runs mostly on US models and resents it, that is dependency with extra steps.

Our read

Two things are true at once. Frontier safety is a genuine question, and the people best placed to answer it are also the people who benefit most from deciding who else may compete. Europe's objection is self-interested too, but at least it says the quiet part out loud.

The encouraging detail is Delangue's split verdict: reject the brake, keep the evaluators. Independent outside access is the one piece of Amodei's plan that does not require trusting incumbents to grade their own homework, and it could survive even if the pacing never happens. If you care about the safety argument rather than the market structure, that is the thread to pull.

What to watch

• Whether the antitrust waiver request attracts any regulator interest, or dies quietly
• Whether Mistral and other European labs publish their own safety proposals rather than objections
• Whether more smaller labs seek embedded-evaluator access, following Hugging Face's lead
• What the Scotland summit produces beyond the group photograph

Discussion spark: Amodei wants a brake and a competition waiver; Mistral calls that a land grab. If the safety case is genuine, how do you slow the frontier without freezing today's pecking order in place?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.