EU AI Act faces uncertain path to global

Anthropic CEO Dario Amodei's weekend call for governments to cooperate on 'pacing the frontier' of AI has intensified debate over global AI governance. His warnings cited risks of losing control over AI systems, cyberattacks, bioterrorism, and major economic disruption.

Kai Zenner, head of the office for European Parliament lawmaker Axel Voss, argues this debate is a perfect opportunity to trigger a 'Brussels effect' for the EU's AI Act. The term describes the EU's ability to project its regulations beyond its borders, as happened with the GDPR data law. Jimmy Farrell, a policy expert at the Brussels think tank Pour Demain, added that Europe could promote being the first major regulator to move on frontier AI risks.

Almost every major AI provider has signed the EU's General-purpose AI (GPAI) Code of Practice. This sets safety obligations for AI model providers and could form the basis for an international governance regime. There are early signs of influence, with U.S. state legislation in California, New York, and Illinois aligning closely with the AI Act's GPAI provisions.

The hurdles to global reach

However, replicating the GDPR's global impact will be difficult for the AI Act. China and the U.S. are developing their own AI approaches, heavily shaped by national security concerns. For these powers, lagging in AI development risks cybersecurity and economic disadvantages. Europe's response to Amodei's catastrophic risk warnings has also been tepid compared to lawmakers in Australia and the U.S.

A European Commission spokesperson reiterated the current focus is on enforcing the AI Act, with developers required to prove their services are safe. More than 30 companies have already received information requests. Although EU tech sovereignty commissioner Henna Virkkunen has called for international cooperation on AI risks, no concrete moves have been announced.

Zenner suggested Brussels may be missing a chance to lead 'a third way' on AI, despite international interest. Potential collaborative work could involve regulatory sandboxes, model evaluation, or incident reporting systems.

Skepticism and regulatory capture

Some policy observers argue that warnings from AI companies do not justify urgent new government action. Patrick Grady, head of European Tech at the Information Technology and Innovation Foundation, said the EU should resist 'scare stories coming out of Silicon Valley, however alluring, and focus on actual adoption risks and rewards.'

Grady added that the warnings could be an attempt at regulatory capture. In this scenario, companies like Anthropic and OpenAI might use fears about AI to influence the terms and teams that regulate their industry.

Enforcement over new interventions

For European lawmaker Brando Benifei, who led negotiations on the AI Act, the immediate priority is clear. 'We cannot calibrate regulation or enforcement to whatever Amodei, [OpenAI's Sam] Altman or [Elon] Musk says on a given day,' Benifei said. He emphasized that Europe has created a legal framework with a public watchdog and enforceable obligations, and using it is the current task.

Benifei noted that international cooperation on frontier AI may resemble the slow process to prevent nuclear proliferation. Ongoing initiatives include the Hiroshima AI Process, which aims to establish international rules for advanced AI, and the OECD's Global Partnership on Artificial Intelligence. Whether this pace is sufficient to prevent the catastrophic risks posed by frontier AI remains unclear.