Altman Describes Three Endgames for AI, Including One Where a Single Firm 'Melts Everyone Else's GPUs'

Stephanie Palazzolo and Amir Efrati report that Sam Altman last week told OpenAI staff about three proposals discussed for safely developing AI (two people with knowledge, plus one with knowledge of specific quotes): an IAEA-style international body, which he appeared to support; a 'Demis plan' merging major labs into one CERN-like entity supplying models to cooperating countries; and an 'Eliezer plan' in which one developer gains enough power to deprive others of compute, which Altman called 'terrible' and listed for completeness (one person saw it as a dig at Anthropic). He said OpenAI must decide within the next year which path to push and can't cede ground to rivals now. Context: OpenAI, Google and Anthropic are working on a Standards Authority for Frontier AI hoped for by end-2026 or early 2027, but disagree on who tests models; Anthropic pitched a draft bill letting Commerce halt models that could cause over $1B in damage or harm more than 50 people; Meta's Zuckerberg and Nvidia's Huang call such rules regulatory capture. Caveat: based on internal remarks relayed by sources, not a public OpenAI position.