The Accord
Humanity & AI Safety Accord
Draft wording. This text is version 1.1 and has not yet completed domain-expert and legal review. It is published so the platform can be tested end to end. Signatures are permanently bound to the exact version and hash shown below, so later edits cannot change what anyone agreed to.
- Version
- 1.1
- Published
- September 22, 2026
- Canonical text hash (SHA-256)
- 68c352635deac3da302cbdcc26d96c17f370c611180e712b2de06ad81ed60821
The Accord
We call on governments to ensure that increasingly powerful artificial intelligence is developed and deployed with appropriate safeguards. We also ask them to work toward making its benefits accessible to people in every nation. We ask nations to establish effective AI safety oversight, cooperate internationally, and work toward common safeguards for advanced AI systems. Decisions about risks that could profoundly affect society should not be left solely to the organisations developing the technology. People, communities and countries deserve a meaningful voice in shaping their future. We call on elected representatives, scientific institutions, technology companies and international organisations to work together toward transparent and accountable AI governance, so that humanity keeps meaningful control over increasingly capable systems.
What you are signing
In short, the Accord asks for international cooperation on AI safety, effective oversight of advanced AI systems, broader access to its benefits, and a meaningful public voice in decisions about societal risks. Its aim is a future in which humanity keeps meaningful control over increasingly capable AI.
What you are not signing
Signing this Accord does not mean you are claiming, or agreeing to, any of the following:
- that you are an AI expert
- that you endorse any specific technical regulation
- that you support a pause on AI development
- that you support a ban on any particular kind of AI
- that you support any political party
- that you agree with every position taken by any AI-safety organisation
You can sign this and still believe AI will be enormously beneficial. You can sign this and disagree with a pause. That is deliberate.
Why this is deliberately broad
Detailed AI law should be written by elected governments and qualified experts, through democratic and international processes, not by a website, and not by asking the public to endorse technical thresholds they have had no opportunity to examine.
What the public can legitimately express is whether they want their governments to take this seriously and to work together on it. That is what this Accord asks, and nothing more.
This is not where the conversation started
Researchers, industry leaders and hundreds of thousands of members of the public have been raising these concerns for years, through the Center for AI Safety, the Future of Life Institute, PauseAI and others.
This Accord builds on that work. It does not claim ownership of it, and it does not count their supporters as its own.