Top AI company leaders — OpenAI's Sam Altman, Anthropic's Dario Amodei and Google DeepMind's Demis Hassabis — used a G7 summit gathering to propose a US-led international forum, or coalition, to set global AI safety standards. They argue the rules should be written by democratic institutions, not AI labs.
AI Governance · OpenAI Proposal
Altman Calls for a US-Led International Forum to Set AI Safety Rules
In the Financial Times, Sam Altman argues that the rules for advanced AI should be written by democratic institutions — not by the labs building the models. Countries and companies would gain access to the technology by following common safety standards.
3
Precedents cited: aviation safety, global finance, atomic energy
IAEA
The nuclear-energy body Altman proposes as a model for AI
2
Participant types: government reps + independent technical experts
How the mechanism works
COUNTRIES
Earn forum membership by complying with safety standards
→
COMPANIES
Gain access to advanced systems via regular certification
→
RESULT
Common safety standards + broad distribution of the technology
The core question: who makes the rules
AI norms should be set by democratic institutions — elected governments and representatives of citizens — not by the commercial judgment of labs.
Independent technical experts would provide impartial analysis of AI capabilities and risks alongside government representatives.
The case for
Setting shared standards early prevents fragmented national restrictions and keeps advanced AI from concentrating in few hands, while democratic bodies hold the rule-making pen.
The skepticism
Critics question whether a company racing to ship the very models that need regulating should be the one proposing the safety framework — citing tension between his safety and rapid-deployment stances.
Moves in parallel with UN discussions on AI governance, the G7 Hiroshima Process, and the safety stances of competitors such as Anthropic. It is a policy framework — no specific model names, benchmarks, pricing, or release timelines.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…