OpenAI CEO Sam Altman described three possible long-term outcomes for artificial intelligence development during an internal company meeting, according to 36Kr, citing The Information. The scenarios range from international regulation to industry consolidation and a single company achieving overwhelming technological dominance.
The first possibility involves creating a global oversight body modeled on the International Atomic Energy Agency. Such an organization could establish common safety requirements, inspect advanced systems and independently evaluate AI developers. Altman and other OpenAI leaders previously discussed international AI governance in a 2023 publication.
The second scenario, which Altman calls the “Demis Plan,” envisions leading AI research organizations operating under one institution, following a model associated with CERN. Shared resources and safety practices could reduce competitive pressure to release products quickly, although concentrating expertise and computing capacity would raise concerns about excessive institutional power.
The third possibility, known as the “Eliezer Plan,” describes one company gaining a decisive lead through recursive self-improvement, in which AI systems help improve their own capabilities. According to the report, this advantage could become difficult for rivals to overcome and, in an extreme case, allow interference with competitors' GPUs. Altman characterized this outcome as deeply undesirable.
36Kr reports that Altman has also warned about losing control of advanced AI and concentrating its power among a small number of organizations. He supports stronger safety measures and independent evaluations, while acknowledging that OpenAI still faces pressure to compete for technological leadership.