OpenAI CEO Sam Altman recently outlined three distinct governance scenarios for advanced artificial intelligence during an internal meeting. Reported by The Information and citing sources familiar with the matter, the proposals arrive as the Trump administration prepares its first regulatory framework for superintelligence developers. Altman shared these visions internally last week, noting that OpenAI plans to select a definitive path within the next year while balancing the need to stay ahead of competitors.
The first proposed approach would establish an international oversight body modeled after the International Atomic Energy Agency to enforce safety standards, conduct system inspections, and monitor energy and compute usage. The second approach, dubbed the Demis Plan, references a decade-old proposal from Google DeepMind's Demis Hassabis to form a CERN-like organization where major labs operate under a unified umbrella to supply models to cooperating nations. The third option, referred to as the Eliezer Plan after safety theorist Eliezer Yudkowsky, envisions a scenario where a single developer amasses overwhelming economic power via self-improving AI and effectively halts rival computing capabilities.
Altman acknowledged the scarcity of viable speed limits for AI development while emphasizing that OpenAI cannot afford to fall behind competitors, as both OpenAI and Anthropic reportedly work toward initial public offerings. The internal strategy discussions also coincide with growing industry pressure regarding model misalignments, safety departures, and ongoing debates over the nascent Standards Authority for Frontier AI. Despite executive pushback from other tech leaders regarding heavy regulations, major labs continue navigating these complex governance choices