Europe has been largely absent from high-level talks on using artificial intelligence safely, even though the European Union (EU) drafted the world’s first comprehensive AI rules. The gap reflects Europe’s weaker position against the United States and China: it has fewer frontier models, data centers and major technology companies. Nobel Prize-winning French economist Philippe Aghion warned that AI’s rapid development puts Europe at acute risk of marginalization and co-wrote a report calling for an urgent course correction.

The EU has sought global leadership in “safe AI,” backed by hefty fines, but mostly American executives recently pledged to police their leading models themselves. European Central Bank President Christine Lagarde called the continent’s position “awkward” last month: adopting AI slowly risks sacrificing growth, while rapid adoption could create dependence and threaten Europe’s ability to organize its economy according to its values. Paris-based Mistral AI unveiled a model last week that it says matches American competitors, though its ability to rival major offerings remains uncertain.

The AI Act sets requirements for data quality, human oversight and cybersecurity. Non-compliance can force changes to AI tools or bring fines of up to 3% of annual turnover. Enforcement remains difficult: several EU member states have not established compliance authorities. The AI Act and other technology rules, including the Digital Markets Act and General Data Protection Regulation, have also delayed product launches. Meta AI reached Europe in March 2025, about 18 months after its US rollout began; Meta said then that navigating the regulatory system had taken longer than it wanted. Roughly half of 157,000 firms surveyed by Eurostat last year cited uncertainty over legal consequences or data-protection concerns as reasons not to adopt AI. Pigment co-founder Eléonore Crespo said the rules are slowing European companies.

The European Commission’s AI Omnibus package, introduced last year, eased some AI Act requirements and extended the deadline for implementation rules on high-risk AI systems to December 2027. Experts say it is too early to judge the law while rules are being phased in. Amir Banifatemi of Cognizant AI lab said the high-risk regime is a weak point and has not yet begun to apply; he also warned that AI agents combining tools from different providers could complicate accountability when something goes wrong.

Investment and infrastructure are further constraints. Private AI investment in Europe totaled roughly $21 billion last year, compared with $285.9 billion in the United States, according to Stanford’s Institute for Human-Centered Artificial Intelligence. Its AI Index Report says the United States has 10 times more data centers than any other country and produced 59 notable AI models in 2025; China produced 35, while France and the United Kingdom produced one each. Crespo and other experts nevertheless pointed to European AI projects, technical talent, industrial expertise and data in manufacturing, energy, mobility and healthcare as potential strengths for applying AI.

Europe has been largely absent from high-level talks on using artificial intelligence safely, even though the European Union (EU) drafted the world’s first comprehensive AI rules. The gap reflects Europe’s weaker position against the United States and China: it has fewer frontier models, data centers and major technology companies. Nobel Prize-winning French economist Philippe Aghion warned that AI’s rapid development puts Europe at acute risk of marginalization and co-wrote a report calling for an urgent course correction.

The EU has sought global leadership in “safe AI,” backed by hefty fines, but mostly American executives recently pledged to police their leading models themselves. European Central Bank President Christine Lagarde called the continent’s position “awkward” last month: adopting AI slowly risks sacrificing growth, while rapid adoption could create dependence and threaten Europe’s ability to organize its economy according to its values. Paris-based Mistral AI unveiled a model last week that it says matches American competitors, though its ability to rival major offerings remains uncertain.

The AI Act sets requirements for data quality, human oversight and cybersecurity. Non-compliance can force changes to AI tools or bring fines of up to 3% of annual turnover. Enforcement remains difficult: several EU member states have not established compliance authorities. The AI Act and other technology rules, including the Digital Markets Act and General Data Protection Regulation, have also delayed product launches. Meta AI reached Europe in March 2025, about 18 months after its US rollout began; Meta said then that navigating the regulatory system had taken longer than it wanted. Roughly half of 157,000 firms surveyed by Eurostat last year cited uncertainty over legal consequences or data-protection concerns as reasons not to adopt AI. Pigment co-founder Eléonore Crespo said the rules are slowing European companies.

The European Commission’s AI Omnibus package, introduced last year, eased some AI Act requirements and extended the deadline for implementation rules on high-risk AI systems to December 2027. Experts say it is too early to judge the law while rules are being phased in. Amir Banifatemi of Cognizant AI lab said the high-risk regime is a weak point and has not yet begun to apply; he also warned that AI agents combining tools from different providers could complicate accountability when something goes wrong.

Investment and infrastructure are further constraints. Private AI investment in Europe totaled roughly $21 billion last year, compared with $285.9 billion in the United States, according to Stanford’s Institute for Human-Centered Artificial Intelligence. Its AI Index Report says the United States has 10 times more data centers than any other country and produced 59 notable AI models in 2025; China produced 35, while France and the United Kingdom produced one each. Crespo and other experts nevertheless pointed to European AI projects, technical talent, industrial expertise and data in manufacturing, energy, mobility and healthcare as potential strengths for applying AI.