Why has Europe gone quiet in the global AI safety debate that America and China keep dominating

[MachineSaint70]

While the United States and China trade increasingly public arguments over AI risk, pacing and regulation, Europe has found itself an oddly muted voice in a conversation it arguably has the strongest regulatory claim to lead. The core problem is structural rather than political, Europe simply does not have anything resembling an OpenAI, an Anthropic or an Nvidia of its own, and that absence of a genuine homegrown frontier lab has quietly diminished the continent's standing in a debate increasingly framed around what the labs actually building these systems say and do.

Christine Lagarde, in her role heading the European Central Bank, lays out the dilemma about as starkly as anyone in European policy circles has managed. Embrace AI wholesale and Europe risks locking itself into permanent dependency on American and Chinese tools and infrastructure, reject it and the continent forfeits whatever growth AI adoption might deliver over the coming decade. Her preferred way out is European self sufficiency, building indigenous AI technology and the datacentre infrastructure to actually run it domestically rather than renting compute and models from elsewhere indefinitely.

Margrethe Vestager, in a co-authored report, frames the stakes in similarly blunt economic terms, warning that Europeans end up bearing the economic consequences of AI decisions made entirely outside their borders regardless of whether they had any say in those decisions. Her emphasis falls on building resilience against what she terms AI crises and developing genuine homegrown capability, while still acknowledging that some of the more existential risks under discussion require solutions that go well beyond just the current US-China rivalry framing that dominates most coverage.

Ursula von der Leyen has tried to stake out a more diplomatic middle position, proposing that Europe invite the major frontier labs to a dedicated discussion specifically about pacing development responsibly. That proposal has gained remarkably little traction so far, especially compared to how forcefully both the Trump administration and Chinese officials have already dismissed any suggestion of a coordinated industry wide slowdown, leaving von der Leyen's invitation looking somewhat sidelined before it even got a proper hearing.

Not everyone accepts the entire framing that Europe is simply behind and needs to catch up though. Frederike Kaltheuner, an adviser at the AI Now Institute, argues the whole behind narrative has been aggressively promoted by American tech executives with an obvious interest in selling that story, and points out plenty of European companies deliberately avoid the most expensive frontier models over legitimate cost and data protection concerns rather than incompetence. She suggests Europe's more cautious regulatory approach through the EU AI Act, which requires high risk AI developers to report safety incidents, may actually prove pragmatic over the long run even if it never becomes a global benchmark given its compliance burden and narrow territorial scope. Itxaso Dominguez de Olazabal from EDRI pushes back even further, arguing that tech industry advocates reflexively blame regulation for competitiveness problems that actually have much deeper structural roots, particularly Europe's existing dependency on American search, chips and datacentre infrastructure that no amount of deregulation alone is going to reverse quickly.

Why Europe has been absent from the great AI safety debatewww.theguardian.com