The latest AI debate at the G7 should be read as a strategic warning for Europe. The issue is not only whether European governments and companies can access advanced U.S. artificial intelligence models. The deeper question is whether Europe can remain competitive, secure and autonomous in a world where AI is becoming geopolitical infrastructure.

At the 2026 G7 summit in Evian-les-Bains, leaders discussed closer coordination on the risks and opportunities of advanced AI. According to Reuters, they also discussed a potential “trusted partners” scheme that could grant selected non-U.S. nations and companies access to advanced U.S. AI models, including systems developed by Anthropic. The debate followed U.S. restrictions on foreign access to some advanced models, driven by national security concerns.

This is a pivotal moment for the European Union. For years, Europe’s main contribution to the global AI conversation has been regulatory leadership. With the EU AI Act, Europe created the first comprehensive legal framework for artificial intelligence worldwide, positioning itself as the global reference point for trustworthy AI.

But the G7 debate exposes a hard truth: regulation alone is not sovereignty.

The EU AI Act is a major achievement. It establishes a risk-based framework for AI systems, bans certain unacceptable uses, creates obligations for high-risk applications and introduces rules for transparency and general-purpose AI models. The Act entered into force on 1 August 2024 and will become fully applicable on 2 August 2026, with specific exceptions and phased timelines.

The Act is especially relevant in the context of frontier AI. Rules for general-purpose AI models became effective in August 2025, and providers of the most advanced models may have additional obligations related to systemic risk. The General-Purpose AI Code of Practice, published in July 2025, gives providers practical guidance on transparency, copyright, safety and security.

This framework can become a European competitive advantage. Trust is not a soft value. In regulated sectors such as finance, healthcare, public services, energy, defence and industrial automation, trust is a market-enabling asset. Companies will not deploy AI at scale if they cannot explain, audit, secure and govern it.

However, legal certainty is only one layer of AI competitiveness. Europe can set world-class rules, but if the most capable models, cloud platforms, chips and deployment infrastructure remain controlled elsewhere, European companies will still operate inside someone else’s strategic perimeter.

That is the sovereignty gap.

The Anthropic/Mythos case shows why this matters. Mythos is described as an advanced cybersecurity model designed to identify flaws in code and strengthen cyber defences. But cybersecurity experts also warn that such capabilities could be used to accelerate attacks against the same systems they are meant to protect.

This dual-use nature changes the policy landscape. Frontier AI will not be governed like ordinary SaaS. It will increasingly be treated like strategic technology: useful for productivity and innovation, but also relevant to cybersecurity, financial stability, critical infrastructure and national security.

For European companies, this creates a new board-level risk. AI procurement is no longer only about accuracy, cost, integration and compliance. It is also about continuity of access.

A company that builds critical workflows on one foreign-controlled frontier model may be creating a hidden dependency. That dependency may not show up in today’s balance sheet, but it can become material overnight if access is restricted by export controls, security decisions or geopolitical tension.

A “trusted partners” framework could be pragmatic in the short term. Europe needs access to the best AI models, especially for cybersecurity, defence, industrial resilience and financial systems. European Commission President Ursula von der Leyen framed this clearly at the G7, saying it is in the mutual interest of the U.S. and the EU that European citizens and companies can safely use the best AI models.

That position is sensible. Europe should not respond with reflexive anti-Americanism or digital isolationism. The EU needs strong partnerships with the U.S., Canada, the UK, Japan and other democratic allies. AI safety, cybersecurity and model governance require international coordination.

But access granted by allies is not the same as autonomous capability.

A trusted-partner regime still implies that access is conditional. It can be negotiated, limited, prioritised or revoked. For governments and enterprises, that means AI dependency becomes a strategic exposure.

Europe needs partnerships, but it also needs leverage. And leverage comes from capability.

The European Commission has started to recognise this. Its AI Continent strategy aims to strengthen Europe’s AI capacity through investment, infrastructure and industrial adoption. The Commission lists €200 billion to boost AI development in Europe, €20 billion to finance up to five AI gigafactories and 19 AI factories to support startups, industry and research.

This is the right direction. The next AI race will not be won only by building better chatbots. It will be won through compute capacity, secure cloud infrastructure, industrial data access, talent density, capital availability and enterprise deployment capability.

Europe’s strongest opportunity is not to copy Silicon Valley model-for-model. Its opportunity is to dominate the trust-intensive deployment layer of AI: regulated industries, public infrastructure, industrial systems, mobility, energy, healthcare, cybersecurity and enterprise transformation.

These are sectors where Europe already has credibility, domain depth and strong institutional frameworks. If the EU can combine AI compliance with operational excellence, it can create a premium category of enterprise AI: systems that are not just powerful, but deployable, auditable and legally robust.

For European companies, the strategic response should be immediate.

First, audit AI dependencies. Which tools, models and workflows rely on non-European infrastructure? Which are business-critical? Which would create operational risk if access changed?

Second, move toward multi-model architectures. Companies should avoid locking critical systems into a single provider. Model-agnostic orchestration, open-source options and European providers should become part of the enterprise AI roadmap.

Third, align AI governance with the EU AI Act early. This is not only a compliance exercise. It is a market-readiness strategy. Companies that build internal AI governance now will move faster when clients, regulators and procurement teams start demanding evidence of safety, transparency and accountability.

Fourth, treat AI as a corporate strategy issue, not an IT experiment. The organisations that win will be those that connect AI adoption with risk management, data strategy, cybersecurity, legal compliance and business model innovation.

The G7 debate is not just another policy story. It is a signal that access to intelligence is becoming a strategic asset.

Europe has built the regulatory layer. It is building the trust layer. Now it must accelerate the capability layer.

The EU AI Act gives Europe a foundation. The General-Purpose AI Code of Practice gives providers a practical compliance pathway. The AI Continent strategy and AI gigafactories agenda point toward the infrastructure Europe needs.

But the execution gap remains.

The next phase of AI will not be won by jurisdictions that only regulate. Nor will it be won by those that only move fast. It will be won by those that combine trust, infrastructure, capital, talent and adoption capacity.

Europe should remain open, cooperative and globally connected. But it must also become harder to switch off.

That is the real lesson of the G7 AI debate: Europe cannot be merely a trusted user of someone else’s AI future. It must become an active builder of its own.