Sam Altman reportedly called one-company control of AI a ‘terrible’ outcome

Sam Altman reportedly called a future in which one AI company controls competitors’ resources “terrible.” An October 24 Moneywise article published by Yahoo Finance relayed that warning, citing The Information. The remarks belong to an internal OpenAI discussion of three governance scenarios: an international watchdog, a shared laboratory, or a dominant developer.

The underlying discussion was already described by The Decoder on October 20, drawing on The Information’s reporting from people familiar with the matter. OpenAI reportedly plans to choose an approach within the next year. Altman also warned that the company cannot afford to lose ground to rivals.

Three places to put authority

1. International enforcement. The first scenario would create a watchdog modeled on the International Atomic Energy Agency. It would enforce safety standards, inspect systems, require audits, and monitor computing capacity and energy use. Such an arrangement would depend on international cooperation, including with China and other countries developing leading AI models.
2. A common laboratory. The second scenario, reportedly called the “Demis Plan,” would bring major AI labs into one entity or under its umbrella. That organization would supply models to cooperating countries. The comparison was to CERN, drawing on a proposal associated with Demis Hassabis, rather than an outside watchdog supervising separate developers.
3. Control through economic dominance. The third scenario, reportedly called the “Eliezer Plan,” imagined a developer gaining extraordinary economic power through self-improving AI—systems increasingly doing the research that produces their successors. It could then withhold computing capacity from others or, at the extreme, shut down their computing capabilities.

Public standards, national authority

Altman reportedly complained internally that too few workable proposals exist for a “speed limit” on AI development. OpenAI’s earlier public standards proposal offers a narrower comparison with the sweeping institutional options described to staff.

In its September 21 statement, Building standards for the next phase of AI, the company advocated international technical standards while preserving national authority. It said those standards would not themselves be licenses, mandatory prerelease reviews, or approval requirements. Governments would decide whether and how to incorporate them into law.

The company proposed common ways to measure autonomous AI research, determine when human review is needed, and classify and report incidents. It also said standards should be developed transparently with outside experts and should not make competition harder for new entrants or developers releasing model weights.