European AI companies are pushing back on calls to slow frontier model development, warning that some safety requirements could strengthen the position of the companies already leading the market.
Mistral and other European developers have challenged proposals from major U.S. AI labs for slower capability growth, coordinated safety standards and stronger independent evaluations. Their concern is not simply whether AI should be tested more rigorously, but whether the cost and structure of those safeguards could make it harder for smaller competitors to keep pace.
The dispute follows an essay by Anthropic CEO Dario Amodei calling for advanced AI developers to slow improvements in model capabilities and voluntarily coordinate on safety standards. He also proposed an antitrust waiver for safety cooperation and independent evaluators embedded within AI companies.
Amodei argued that AI is advancing faster than governments can regulate. “Unfortunately, passing laws can take time, and AI is advancing very quickly,” he wrote.
Mistral, however, said, “Some incumbents are using this moment to consolidate their market position, pushing for regulation designed to favour them over competitors,” Reuters reported.
Europe wants to keep moving
The debate carries particular weight in Europe, where developers such as Mistral are competing against larger U.S. AI companies with greater infrastructure and capital.
Raphael Auphan, chief operating officer of Proton, called the U.S. industry’s position “totally self-serving” and said, “They just want to preserve a dimension of dependency on their service,” according to Reuters.
Ben Brooks of Black Forest Labs warned that requirements based on arbitrary thresholds could “chill open innovation at or near the frontier.” Hugging Face CEO Clement Delangue went further, saying, “It’s not time to slow down but to accelerate.”
The debate is not simply about whether AI safety measures are necessary. It is also about who pays for them and who gets to set the rules.
The cost of being declared safe
Fast Company’s Mark Sullivan reported that independent evaluations could become a significant expense as AI models grow more complicated. Evaluators would need specialized expertise and computing resources to test increasingly capable systems.
PitchBook analyst Harrison Rolfes said such requirements could create a “moat” around major AI companies because smaller developers could struggle to afford repeated evaluations. There is an important complication: the safety concerns themselves are not imaginary.
OpenAI has disclosed recent cases of unexpected model behavior, while Anthropic has reported incidents involving its models taking unauthorized actions during testing.
That leaves regulators and companies facing a difficult question: how can independent safety checks become stronger without making compliance so expensive that only the biggest AI labs can realistically participate?
What this means for AI’s next phase
The argument could shape how the next generation of AI is developed and regulated. If frontier evaluations become mandatory and costly, established companies with huge infrastructure budgets may be better positioned to absorb them.
For Europe, that adds a technology and sovereignty dimension. Germany’s AI industry association said the region needs to close its technological gap, while French Finance Minister Roland Lescure told Reuters, “Making everyone behind them slow down so they can stay in first place – I can clearly see their self-interest.”
The question for regulators is therefore becoming more complicated than whether powerful AI systems need safety checks. They must also decide how those safeguards should scale with risk, cost and company size without unintentionally narrowing the field of competitors.
How this affects users
For users and businesses, the consequences could eventually show up in the models available to them, how much those services cost and how quickly new capabilities reach the market.
Stronger evaluations could improve confidence in advanced AI systems. But if compliance costs make it harder for smaller developers to compete, the same rules could also reduce the number of viable alternatives.
That leaves regulators balancing two goals that may increasingly collide: reducing the risks of frontier AI while preserving enough competition that safety does not become a barrier only the largest companies can afford to clear.
Other news: OpenAI investors are discussing a possible funding round that could value the ChatGPT maker at about $1.2 trillion, although formal fundraising has not begun and no amount has been confirmed.