Sam Altman called for shared rules for the most capable AI systems at the UN Security Council on September 23, 2026. OpenAI's chief executive identified loss of human control and concentrated power as major risks. He said the company had slowed down unilaterally before and would do so again if necessary.
Altman proposed common standards for capabilities and safeguards, alongside reliable incident-reporting channels. He said further training should require a strong case that humans will remain in control. Governments would decide which rules become law.
Yoshua Bengio argued for stronger requirements at the same meeting. The computer scientist called for independent safety evidence before training and deployment, licensing and liability insurance. His published speech also stresses the scientific independence of assessments.
The National also reported Dario Amodei's video participation and his call for a manageable pace of development. These positions do not amount to an agreed training halt. The speeches reviewed establish demands and commitments, not new binding UN rules or another technical incident.