Anthropic and Google CEOs Brief G7 on AI: US Wants to Lead Global AI Rules

Two AI rivals walked into a G7 summit in France, and the real story isn’t about either of them. Dario Amodei, CEO of Anthropic, and Demis Hassabis, CEO of Google DeepMind, held a closed-door session with G7 leaders to pitch a US-led framework for coordinating global AI development. The presentation wasn’t about their products or their rivalry — it was about convincing the world’s most powerful democracies that American AI companies should shape the rules of the road before China or the EU do it for them.

The Anthropic Google G7 AI session represents a pivotal moment in AI governance. For the first time, competing US AI companies are jointly lobbying foreign governments for a regulatory framework that aligns with American interests. It’s diplomacy by technology, and the stakes couldn’t be higher.

Anthropic Google G7 AI: The Closed-Door Session

The G7 summit in Annecy, France, included an unprecedented technology session on June 15, 2026. Unlike previous G7 discussions about AI — which typically featured government officials talking about AI companies — this session featured AI company leaders talking to government leaders directly.

Dario Amodei and Demis Hassabis presented jointly to representatives from the United States, United Kingdom, Canada, France, Germany, Italy, Japan, and the European Union. The session was closed to press, but multiple diplomatic sources have confirmed the key themes:

• The need for international coordination on AI safety standards and testing requirements
• A proposed US-led coalition to establish shared benchmarks for evaluating AI system capabilities and risks
• Concerns about regulatory fragmentation — specifically, the risk that different nations implementing incompatible AI regulations would slow innovation without improving safety
• The urgency of acting before China establishes its own global AI governance framework through Belt and Road digital infrastructure partnerships

The session lasted approximately 90 minutes — far longer than the 30 minutes originally allocated on the schedule. That extension tells you something about how seriously G7 leaders are taking the AI governance question. This isn’t abstract policy anymore. It’s a strategic priority that’s reshaping international relations.

Why Amodei and Hassabis Presented Together

Anthropic and Google DeepMind are competitors. They compete for talent, for customers, for benchmark leadership, and for the attention of policymakers. The fact that Amodei and Hassabis presented jointly at the G7 is significant — and strategic.

The calculation is straightforward: the competition between US AI companies matters less than the competition between the US and China for AI supremacy. By presenting a unified front, Amodei and Hassabis signal that American AI companies can coordinate on governance even while competing commercially. This is the message G7 leaders need to hear if they’re going to trust US companies to help shape global AI rules.

The partnership also leverages complementary strengths. Anthropic has positioned itself as the “safety-first” AI company — Claude’s development philosophy emphasizes responsible AI development, constitutional AI, and harm reduction. Google DeepMind brings scientific credibility — Gemini’s launch and DeepMind’s Nobel Prize-winning work on protein structure prediction give Hassabis unique authority when discussing AI capabilities and risks.

Together, they present a narrative that US AI development is both capable and responsible — exactly the message needed to counter concerns about unregulated AI development. Whether that narrative is entirely accurate is another question, but in diplomatic contexts, perception matters as much as reality.

The US Bid for Global AI Governance

The Anthropic Google G7 AI session is part of a broader US strategy to shape global AI governance before other power centers establish competing frameworks.

The US approach, as articulated by Amodei and Hassabis, centers on several principles:

Voluntary industry commitments with government oversight: Rather than top-down regulation (the EU approach), the US favors industry-led safety standards with government monitoring and enforcement. This is essentially the model that governed early internet development — industry sets technical standards, government provides a legal framework.

International safety testing collaboration: A shared set of benchmarks and evaluation criteria for AI systems, administered by national AI safety institutes (like the UK AISI and US NIST) that share results and methodologies. This would create a common language for assessing AI risks across borders.

Open research sharing on safety: Encouraging AI companies to publish safety research while protecting competitive intellectual property. Both Anthropic and Google DeepMind already publish extensively on AI safety topics — this framework would formalize and incentivize that practice across the industry.

Export control coordination: Aligning chip and technology export controls across G7 nations to prevent circumvention. Current export controls are primarily US-driven, and gaps in allied nations’ regulations create loopholes that undermine effectiveness.

The underlying message is clear: the US wants to write the rules for global AI governance, and it wants American companies — specifically safety-focused companies like Anthropic and research-driven companies like Google DeepMind — to be at the table when those rules are written.

China and EU: The Alternative AI Frameworks

The urgency behind the Anthropic Google G7 AI presentation stems from two competing governance frameworks that are already taking shape:

China’s approach: China has been building AI governance influence through the Global AI Governance Initiative, proposed in 2023 and expanded since. Through Belt and Road digital infrastructure partnerships, China is offering developing nations AI tools, training, and infrastructure in exchange for alignment with Chinese technical standards and governance norms. Countries that adopt Chinese AI infrastructure will naturally align with Chinese regulatory frameworks — creating a governance bloc that could rival Western-led initiatives.

EU’s approach: The EU AI Act, which entered full enforcement in 2025, takes a prescriptive regulatory approach that US companies have criticized as overly burdensome. It categorizes AI systems by risk level and imposes specific requirements on each category — an approach that Silicon Valley argues stifles innovation and gives non-EU competitors an advantage.

The US fear — articulated by both Amodei and Hassabis — is that regulatory fragmentation could create a world where AI companies must comply with three or four incompatible regulatory regimes, slowing development without improving safety. By proposing a coordinated G7 framework, the US hopes to establish a unified standard that allies adopt, reducing compliance burdens while maintaining meaningful safety requirements.

There’s also a competitive dimension. If the EU’s strict regulatory framework becomes the global standard, European regulators effectively gain veto power over AI capabilities — something US companies want to avoid. And if China’s governance model spreads through Belt and Road partnerships, US strategic interests in AI leadership are directly threatened.

What Anthropic and Google Actually Proposed at G7

Based on diplomatic sources and subsequent public statements, the Amodei-Hassabis presentation included several specific proposals:

1. G7 AI Safety Testing Protocol: A standardized framework for evaluating frontier AI models before deployment, with shared benchmarks across G7 nations. This would include mandatory red-teaming for models above certain capability thresholds and shared databases of discovered vulnerabilities.

2. Compute Governance Framework: A coordinated approach to monitoring and governing the large-scale compute infrastructure used for AI training. This includes sharing information about training runs above certain FLOP thresholds and coordinating export controls on AI-relevant hardware.

3. AI Incident Reporting System: A shared platform for reporting and analyzing AI-related incidents — similar to how aviation has the FAA’s incident reporting system. When an AI system causes harm or exhibits unexpected behavior, the incident would be reported and analyzed collaboratively across G7 nations.

4. Research Collaboration on AI Safety: Formal partnerships between national AI safety institutes (US NIST, UK AISI, EU AIA, Japan AISI) to share safety research, evaluation methodologies, and risk assessments. This would include joint funding for safety research and shared access to evaluation infrastructure.

These proposals are notable for what they include — and what they omit. There’s no mention of mandatory licensing for AI models, no proposal for restricting open-source AI development, and no call for pre-deployment government approval. This is a “light touch” governance framework designed to be acceptable to the US tech industry while providing enough substance to satisfy G7 allies who want meaningful AI oversight.

The OpenAI Absence Is Telling

Perhaps the most notable aspect of the Anthropic Google G7 AI session is who wasn’t there: Sam Altman and OpenAI.

OpenAI is the largest AI company by valuation, the most widely used AI product by consumer reach, and the most politically connected AI company in Washington. Its absence from the G7 presentation is conspicuous and reveals fractures within the US AI industry’s approach to governance.

Several factors likely contributed to OpenAI’s exclusion:

Commercial focus vs. safety credibility: OpenAI has moved aggressively toward commercialization — the AI phone project, enterprise contracts, and now advertising. This commercial aggressiveness undermines the “responsible development” narrative that’s essential for credibility in governance discussions. Anthropic and Google DeepMind, by contrast, maintain stronger safety research profiles.

Governance positioning: OpenAI has historically pushed for less regulation, arguing that AI development should move fast and that governance should follow capability, not precede it. This position is harder to sell to G7 leaders who are increasingly concerned about AI risks.

The Altman factor: Sam Altman’s high profile and controversial leadership style — including his temporary ouster and return in late 2023 — make him a less ideal representative for a diplomatic presentation focused on stability and responsibility. Amodei and Hassabis project a more measured, academic authority that plays better in government settings.

The absence also reflects a genuine strategic divergence within US AI. OpenAI wants to build products and capture market share. Anthropic wants to shape governance frameworks. Google wants to maintain its infrastructure advantages. At the G7, the governance-focused approach won the invitation.

What the Anthropic Google G7 AI Summit Actually Means

The Anthropic Google G7 AI session matters for three reasons:

First, AI governance is now a top-tier diplomatic issue. When AI company CEOs get 90 minutes with G7 leaders, it signals that AI governance has moved from the technology policy backwater to the main stage of international relations. This is no longer a niche issue for tech regulators — it’s a strategic priority alongside trade, defense, and climate.

Second, the US is playing to win the governance race. By sending Amodei and Hassabis to pitch a US-led framework, Washington is making a clear play to set the global standard for AI governance before competitors can. The alternative — letting the EU’s precautionary approach or China’s state-controlled model become the default — is seen as unacceptable to US strategic interests.

Third, corporate influence on AI governance is accelerating. Having AI company CEOs directly brief world leaders on proposed governance frameworks blurs the line between corporate lobbying and diplomatic engagement. Whether this is good governance or regulatory capture depends on your perspective — and probably on whether the resulting frameworks actually protect public interests or primarily protect corporate ones.

The rapid evolution of AI capabilities means governance frameworks created today will shape AI development for decades. The decisions made in rooms like the one at the G7 summit — by people like Amodei and Hassabis — will determine whether AI development proceeds with meaningful safety guardrails or whether “governance” becomes a euphemism for industry self-regulation.

Two AI rivals walked into a G7 summit together. The real competition isn’t between Anthropic and Google — it’s between the US and China for the right to define how the most powerful technology in human history is governed. The future of AI depends on who wins that competition, and whether the winning framework prioritizes human interests over corporate and national ones.

Based on what we’ve seen so far, I wouldn’t bet on the humans.