Major European artificial intelligence (AI) companies are pushing back against calls from leading U.S. AI labs to slow frontier AI development. With Europe already behind the United States and China in model competitiveness, computing infrastructure and investment scale, they worry that any slowdown would further cement U.S. dominance.

On Sept. 18, blockchain media outlet Cryptopolitan reported that European AI firms including France's Mistral voiced concern that a recently floated plan to "slow frontier AI development" could hinder competition.

The controversy began with an essay titled "We should regulate the frontier" that Anthropic Chief Executive Dario Amodei released on Sept. 12. Amodei proposed slowing the pace of AI capability advances while continuing product releases and model training.

He outlined a three-step approach: granting external evaluators access similar to that of AI lab staff, cooperation among AI labs in democratic countries, and eventually expanding into an international cooperation framework. The aim is to secure safety while managing the speed of AI performance gains.

European AI firms said the discussion could affect market competition beyond AI safety. Mistral raised concern that incumbent leaders could push for rules that favour themselves over competitors.

Proton Chief Operating Officer Raphael Auphan (라파엘 오팡) criticised the move as "entirely self-interested" and argued it aims to maintain dependence on U.S. companies. Black Forest Labs also said applying unclear standards could block innovation and prevent new firms from entering the market.

Europe's pushback reflects its AI industry gap with the United States. Prosus data showed Europe accounted for 11 percent of global AI venture investment in 2023, while the United States took 77 percent. The gap widened further in later funding rounds.

Differences also appear in AI model development. European Central Bank President Christine Lagarde said in a Sept. 14 speech that, as of 2025, there were 59 notable AI models from the United States and 35 from China, while France and Britain each had 1.

Global AI data centre computing power is also concentrated in the United States. Lagarde said the United States accounts for about 75 percent of global AI data centre computing power, while Europe is around 5 percent.

The model performance gap is also fuelling European concern. In Epoch AI's ECI ranking, the U.S. GPT-6 Astra led with 166 points, and China's Kimi K3 ranked 11th with 158 points. Mistral Medium 3.5, which France's Mistral unveiled in April, scored 142.5 points.

European companies say additional constraints on development speed would make it harder to narrow the gap as they try to catch up with the United States and China. They also worry that new evaluation standards or international cooperation frameworks could impose higher costs on small and mid-sized European developers with fewer resources than large AI companies.

Europe does not lag the United States in every area of AI use. Microsoft's global AI diffusion report showed generative AI usage rates, measured by working-age population, were 48.6 percent in Norway, 48.4 percent in Ireland and 47.8 percent in France, higher than 31.3 percent in the United States. Europe trails in cutting-edge models and computing infrastructure, but European countries are seeing rapid uptake in real-world AI use.

The European Union is also investing to strengthen AI industry competitiveness. The EU's "AI Continent Action Plan" includes a plan to mobilise 200 billion euros for AI. It includes "InvestAI" of 20 billion euros to support up to 5 AI gigafactories and plans to build at least 19 AI factories.

The EU is also pushing to expand AI adoption. The European Commission puts AI adoption among EU companies at 13.5 percent and is pursuing an "Apply AI" strategy to raise it.

Views in the U.S. AI industry are split on Amodei's proposal. Elon Musk and Sam Altman expressed sympathy with the idea of bringing in external evaluators. Altman also said slowing the pace should not mean halting AI development itself.

Hugging Face's Clement Delangue also stressed that Europe must keep moving forward in AI development while agreeing on the need for an independent evaluation framework.

Ultimately, the core of the debate is not simply choosing between AI safety and development speed. Who sets the rules for AI development, and who bears the costs when new rules are introduced, are emerging as key issues.

The Organisation for Economic Co-operation and Development (OECD) analysed that as innovation becomes more concentrated in the AI industry, revenue concentration also tends to rise. With AI startups that received venture investment continuing to be acquired by large incumbents, small and mid-sized European developers worry that new evaluation and cooperation frameworks could become another barrier to entry.

The controversy also affected financial markets. On Sept. 14, shares of major AI companies fell, showing that debate over development speed and regulation can affect investor sentiment as well as the competitive landscape in the technology industry.

European AI firms' pushback is less an objection to AI safety regulation itself than a challenge over how new rules limiting development speed could affect competition as Europe tries to catch up with the United States and China. How to coordinate international cooperation for AI safety with competition among countries and companies is expected to become a key task.