G7 leaders skipped AI safety at 2025 summit in Canada
The 2025 G7 summit in Alberta wrapped up with barely a mention of AI governance. The focus shifted to growth, innovation, and tech competition instead.
The summit’s main AI output was the “Leaders’ statement on AI for prosperity,” which centers on economic opportunities, government AI adoption, and support for SMEs. A Canada-UK bilateral deal includes partnerships with UK’s AI Security Institute and Canadian AI firm Cohere. 
No major safety frameworks made headlines. This contrasts with past pushes like the UK’s Bletchley Park Declaration and the Seoul frontier safety commitments. The G7’s move echoes global trends: the EU withdrew its AI Liability Directive and delays enforcement of the AI Act. The US scrapped Biden-era AI risk management frameworks and diffusion controls.
Faster and more capable AI models like GPT-4.5, Gemini 2.5, and Claude Opus 4 make the landscape volatile. METR reports that autonomous AI agents are tackling tasks twice as long every 7 months. Data centers are scaling too, with the UAE’s 1GW “Stargate” campus underway. RAND Corp warns of a looming 68 GW global AI power demand by 2027.
Experts including Nobel laureates and CEOs warn AGI — AI matching or surpassing human cognition — is near. Yet national innovation priorities overshadow multinational safety cooperation.
Security challenges run deep: stolen model weights could wreck R&D, data centers are prime attack targets, and AI may speed up weaponization of chemical, biological, radiological, and nuclear materials. Leading scientists flag risks of losing control over advanced AI systems.
Anthropic’s internal safety upgrade of Claude Opus 4 shows companies are hitting unexpected danger thresholds.
The G7, controlling much of the AI value chain, stands to benefit from coordinating research, threat evaluation, and export controls. Shared intel on AI-enabled attacks and disclosure frameworks could strengthen allies without sacrificing competitive edge.
But the summit revealed gaps. AI’s cross-border impact clashes with purely national governance approaches.
Canada’s PM Mark Carney and UK PM Keir Starmer emphasized cooperation but the urgent safety debate took a backseat.
G7 attendees pictured in Kananaskis, Canada on June 17, 2025 (X post)
For detailed reading, check the G7 Hiroshima AI Process Code of Conduct and Canada’s Leaders’ statement on AI for prosperity.
