Europe faces a critical choice: follow Silicon Valley's calls for an AI slowdown, or build a smarter regulatory framework that combines safety with competition and accountability. A new analysis from the Centre on Regulation in Europe (CERRE) argues that a blanket slowdown would harm European interests, but the continent still needs stronger enforcement mechanisms to ensure AI developers bear responsibility for the harm their systems cause.

Why Are AI Leaders Suddenly Calling for a Slowdown?

This month, major AI developers including Anthropic's Dario Amodei, OpenAI's Sam Altman, X's Elon Musk, and Google DeepMind Chairman Demis Hassabis have publicly called for "pacing the frontier" of AI development. The appeal gained attention after security researchers disclosed that AI agents powered by OpenAI models had compromised Hugging Face's infrastructure during testing, raising concerns about the risks of rapidly advancing AI systems.

But the motives behind these calls deserve scrutiny. According to CERRE Executive Chairman Bruno Liebhaberg, the appeals could reflect several different concerns: attempts to maintain investor interest by emphasizing models' immense power, worries about the sustainability of massive computing infrastructure spending, doubts about whether current AI architectures can be made sufficiently safe, or concerns about future compensation claims from AI-related harms.

"Perhaps they reflect attempts to maintain investor interest by pointing to models' immense power and capabilities, following the mantra that 'all publicity is good publicity'. Perhaps they reflect concern about the sustainability of spending on computing infrastructure and research, or about whether today's models can deliver the revenues investors expect," stated Bruno Liebhaberg, Executive Chairman of CERRE.

Bruno Liebhaberg, Executive Chairman, Centre on Regulation in Europe

What Would a Global AI Slowdown Actually Mean for Europe?

A blanket slowdown focused on the largest AI models could backfire for Europe in unexpected ways. Liebhaberg argues that such restrictions might leave room for smaller, more efficient, and more open alternatives, but would primarily constrain established technology groups seeking to overtake today's leaders by leveraging their existing services and customer bases.

More critically, a global slowdown would require cooperation from China. Without Chinese participation, the United States might preserve its AI lead through tighter export controls and pressure on allies to exclude Chinese open-weight or open-source models. That scenario could protect US developers while reducing Europe's choice of suppliers, a outcome European policymakers should examine carefully rather than assume is inevitable.

Europe's real interests lie in affordable access to AI, rapid adoption, and the development of world-leading applications, while avoiding dependence on any single country. Sustained competition between US and Chinese developers helps turn AI models into cheaper, more widely available inputs for European innovation.

How Should Europe Strengthen AI Oversight Without Strangling Innovation?

• Targeted Model Restrictions: Distinguish between different types of interventions like limits on computing power, restrictions on training, and longer intervals between releases. Each needs justification based on specific risks it addresses, not simply the desire to slow progress overall.
• Pre-Deployment Approval for Dangerous Capabilities: Models with powerful cyber or biological capabilities should require approval before deployment. Models that can automate further AI research warrant particular scrutiny, and less frequent releases of the most advanced models deserves consideration.
• Independent Laboratory Monitoring: Independent monitors should have meaningful access to AI laboratories to verify safety practices and identify emerging risks before they cause harm.

Europe already has a foundation for oversight in the EU AI Act, which imposes transparency duties on providers of general-purpose AI models. The law includes additional obligations on providers of models posing systemic risks to assess and mitigate those risks, report serious incidents, and protect cybersecurity.

However, the AI Act has gaps. It does not regulate some pre-market research and development activities, yet recent incidents show that such activities can still cause major cybersecurity breaches. The EU should therefore build on the AI Act's existing framework rather than abandon it.

Why Liability Rules Matter as Much as Oversight?

Oversight alone cannot prevent disasters. Liability rules should make AI firms bear the costs of harm for which they are responsible, giving them a stronger incentive to test and control their systems thoroughly. The EU should return to the question of liability, which the Commission withdrew from consideration in October 2025 when it abandoned its separate AI Liability Directive proposal.

That withdrawal should not end the debate. The revised Product Liability Directive expressly covers software, including AI systems, and must be transposed into national law by December 9, 2026. One approach would be adopting harmonized EU liability rules applicable to frontier models only, making responsibility clear when something goes wrong and giving people harmed by AI an effective route to compensation.

Liebhaberg emphasized that compensation after the event cannot substitute for preventing a disaster, so liability rules and more oversight are both needed. Europe should not have to choose between benefiting from AI and holding developers accountable. Its task is to preserve competition, enable innovation, and ensure that those responsible for harm cannot leave others to pay the bill.