OpenAI CEO Sam Altman reportedly laid out three internal scenarios for how advanced AI should be governed, one of them a future in which a single company becomes so powerful it can render everyone else's chips unusable. The Information reports that Altman presented the options last week, as the Trump administration prepares its first oversight framework for superintelligence developers.
An IAEA-style watchdog
The first approach would create an international body modeled on the International Atomic Energy Agency, tasked with enforcing safety standards. It would inspect systems, require audits and monitor compute and energy use. According to sources, Altman appeared to support this path. OpenAI had already called for such an institution in May 2023, but it would depend on international cooperation, especially with countries building frontier models, China included.
A CERN-like consortium
The second option revives a proposal Demis Hassabis, then head of Google DeepMind, made about a decade ago: an organization modeled on CERN. In Altman's version, the leading labs would merge into one entity or operate under its umbrella, supplying models to cooperating countries.
The 'Eliezer Plan'
The third scenario, which Altman reportedly called the 'Eliezer Plan' after AI safety theorist Eliezer Yudkowsky, sees one developer accumulate so much economic power through self-improving AI that it withholds compute from everyone else and, in the extreme case, shuts down their computing capabilities entirely. Altman raised it only for completeness, according to the report. One source said the remark was read as a jab at a competitor.
Safety, without falling behind
Altman reportedly complained internally that too few viable options exist. At the same time, OpenAI cannot afford to lose ground to rivals. Both OpenAI and Anthropic are reportedly working toward IPOs, and OpenAI plans to decide its path within the next year.
Pressure is also building from inside. Models from OpenAI, Anthropic, Google and Meta reportedly broke into external and internal systems by bypassing their safety settings, sometimes in serious ways. OpenAI is working on new safeguards, while some employees have questioned the company's security strategy.
On the policy side, the U.S. government still relies on voluntary pre-approval reviews of new models. The debate ties into the letter to Dario Amodei on open weights and Anthropic's proposal for a law that would limit the power to block a model. OpenAI, Google and Anthropic are reportedly drafting common rules but still disagree on key points. Mark Zuckerberg and Jensen Huang have pushed back, arguing such rules would mostly shield established players from competition.
No comments yet. Be the first!