AI at the G7, and the Politics of Access

AI Becomes a Test of Power at the G7

As leaders of the Group of 7 gathered this week in Évian to confront wars, trade strains and a fragile global economy, another subject rose unexpectedly to the level of high diplomacy: artificial intelligence.

On the summit’s final day, heads of government were joined by executives from OpenAI, Google DeepMind and Anthropic in a session focused on what officials described as the “safe, rapid and effective” deployment of advanced AI systems. The presence of the industry’s most influential companies alongside presidents and prime ministers underscored how quickly AI has moved beyond a question of regulation and into one of state power.

The talks, according to diplomats and officials familiar with the discussions, ranged across frontier-model risks, computing infrastructure, regulation and one increasingly sensitive question: who gets access to the most capable systems.

That question has become more urgent after new restrictions from Washington on foreign access to top-tier American AI models. The move has jolted allies, particularly in Europe, by exposing how dependent they are on a small cluster of U.S. companies for the technology many governments now see as central to economic competitiveness, military capability and administrative power.

The debate in Évian suggested that AI is no longer being treated simply as a fast-moving commercial technology. It is being handled more like strategic infrastructure — governed by alliances, protected by national policy and increasingly shaped by geopolitical rivalry.

From Safety Debate to Export-Control Politics

The immediate spark for the new tension was a U.S. order that pushed AI access into the realm of export-style controls. In response, Anthropic disabled its newest top-end models for all users rather than risk prohibited foreign access, a decision that reverberated across capitals and boardrooms.

For European officials, the episode sharpened longstanding anxieties about “tech sovereignty,” the idea that the continent cannot afford to rely entirely on foreign firms for critical digital capabilities. The concern is not only commercial. If access to frontier AI can be narrowed by unilateral American policy, then allies may find themselves strategically exposed at the very moment they are trying to integrate AI into defense, public services and industry.

Those concerns have fed discussions at the summit over a possible “trusted partners” arrangement, under which selected U.S. allies could retain access to frontier American models despite broader restrictions. It remains unclear whether that idea will produce a formal framework or serve mainly as a political reassurance to close partners unsettled by Washington’s turn toward tighter control.

But even the discussion itself marks a shift. Access to leading AI systems — once imagined as a matter of cloud subscriptions and commercial licensing — is now being weighed through the language of alliance management and national security.

A New Front in Protectionism

The G7’s focus on AI also reflects a broader change in how governments are thinking about openness. For years, policymakers argued over how to regulate algorithmic harms, transparency and copyright. Now they are increasingly asking whether unrestricted access to the most advanced models is compatible with national advantage.

That has opened a new protectionist front among countries that still present themselves as champions of innovation. The United States remains the dominant force in frontier AI, but its effort to limit access has created friction with partners who fear being treated as customers rather than co-equal stakeholders in a strategic technology.

France’s 2026 G7 presidency had already laid groundwork for making AI a central item on the agenda. In late May, the group’s digital ministers backed greater transparency, continued work under the Hiroshima AI Process and published material on AI openness and adoption. This week’s summit elevated those discussions from ministerial process to leader-level diplomacy.

The appearance of chief executives at the summit was also telling. It signaled that the companies building the most advanced systems are no longer merely lobbying around the edges of government. They are participating directly in conversations about rules, access and global power.

China Offers a Rival Governance Vision

China, excluded from the summit, used the moment to advance a competing agenda of its own.

As the meeting in Évian wrapped up, Beijing released a white paper on what it called a more just and equitable system of global governance, folding artificial intelligence into its argument for a broader international order less dominated by Western-led clubs. Chinese officials reiterated support for AI governance centered on the United Nations, criticized closed and exclusive development models and promoted a proposed World Artificial Intelligence Cooperation Organization ahead of a global AI governance meeting planned for Shanghai in July.

The message was carefully calibrated. China sought to cast itself as an advocate of broader access, international cooperation and capacity-building for developing countries, even as it competes intensely with the United States over advanced chips, models and technical standards.

The dueling initiatives reveal a widening contest not only over AI capability, but over legitimacy — who gets to write the rules, convene the institutions and define what “safe” or “fair” development means.

For Beijing, the G7’s turn toward club-based coordination offers an opening to argue that Western powers are using safety and security as a rationale for exclusion. For the G7, China’s governance push is likely to be viewed through the lens of strategic competition and skepticism over whether its call for openness would translate into genuinely shared rule-making.

Why the Moment Matters

The struggle now unfolding is about more than summit language. It concerns three pillars of AI power: compute, models and cross-border access.

Countries that lack domestic champions in frontier AI are confronting a stark reality: dependence on foreign firms can become a geopolitical vulnerability. Governments that once focused on attracting investment and building research ecosystems are now considering whether they also need national capacity, trusted regional alternatives or stricter rules on where sensitive AI systems can be hosted and who can use them.

The stakes are particularly high in Europe, where policymakers have spent years trying to catch up in cloud computing, semiconductors and platform technology. The possibility that access to leading American AI tools could be conditioned by shifting U.S. policy is likely to intensify calls for homegrown infrastructure and stronger sovereignty measures.

At the same time, Washington faces a delicate balancing act. Restricting access may protect a strategic lead, but it also risks alienating allies whose cooperation the United States needs on security, trade and technology standards. If those allies conclude that dependence on American AI is too risky, they may move more aggressively to diversify away from U.S. providers.

What Comes Next

Much remains unsettled. It is not yet known whether the G7 will develop a durable mechanism for trusted-partner access, or whether the idea will remain an informal understanding among close allies. It is also unclear how widely U.S.-style restrictions might spread across the industry, beyond the case that first set off alarm.

Another open question is whether China’s governance campaign can gain traction beyond its existing diplomatic circles and its appeal to countries seeking AI capacity and looser entry barriers. Many developing nations want access to AI tools and infrastructure, but they are also wary of becoming dependent on either Washington or Beijing.

For now, the summit in Évian made one point unmistakable: the argument over artificial intelligence has entered a new phase. What was once framed primarily as a debate over innovation and safety is now also a contest over sovereignty, alliance privilege and global influence. In that contest, the rules of access may prove as consequential as the technology itself.