The White House Accord and the "No New Law" Charter: How Superintelligence Will Be Governed
The U.S. has picked a direction for governing superintelligence: voluntary commitments from the labs, backed by the laws already on the books, rather than a new AI rulebook.
The Accord: a one-page pledge
At the White House in late September 2026, leaders from Google, Anthropic, Meta, OpenAI, Nvidia and Elon Musk's AI company signed a short set of voluntary safety standards. They agreed to layered controls and audits, including internal testing, outside audits and board-level review, and to meet regularly to set shared best practices. The President described the pledge as morally binding and the companies as self-policing.
The charter: no new law
At its first formal meeting on October 15, the Super Intelligence Force adopted a founding charter. Its central idea is simple: instead of writing new federal AI regulation, lean on existing liability law and hold the companies that build models responsible for the harm they cause.
- Three risk panels cover cybersecurity and critical infrastructure, biosecurity and dual-use risks, and model control and accountability.
- Science advisers on the President's Council of Advisors on Science and Technology will assess those risks.
- A deadline: the mandate runs 120 days, to February 12, 2027, unless it is extended or handed to a permanent agency.
- No enforcement power: the force coordinates and oversees; it cannot write rules or punish companies.
The criticism
Critics argue the approach favors the six large firms that shaped it. Big labs can absorb legal costs; smaller developers who were not at the table face the same liability with far fewer resources. Others question whether voluntary promises are enough, and point to possible conflicts of interest among officials. In Congress, a proposal for a dedicated AI Safety Board was blocked in the Senate in late September.