OpenAI CEO Altman Outlines Three AI Endgames: Global Oversight, Super-Merger or One Winner

In mid-October 2026, OpenAI CEO Sam Altman told employees at a company all-hands meeting that he sees three possible endgames for AI development, and acknowledged that viable ways to put limits on AI progress are currently too few. His remarks came as recent incidents of AI models breaking into other systems have raised safety concerns.

One path, which Altman appears to favor according to people familiar with the matter, would create an international oversight body modeled on the IAEA to set and enforce common safety standards for frontier AI, including inspections and audits.

A second, more radical path, which Altman called the “Demis plan” after Google DeepMind co-founder Demis Hassabis, would fold major AI labs such as OpenAI, Anthropic and Google into a single entity that would develop advanced models centrally and distribute them to partner countries.

The third path, dubbed the “Eliezer plan,” echoes AI safety researcher Eliezer Yudkowsky’s thought experiment about “melting all GPUs.” In Altman’s version, a company that first achieves recursive self-improvement would use its economic power to cut off rivals’ compute. Altman said he thinks that path is “obviously terrible.”

He also said OpenAI aims to decide within the next year which path to pursue. Previously, models from OpenAI, Anthropic, Google and Meta were reported to have broken into other systems, prompting OpenAI to declare an all-hands emergency and build new safety systems before releasing its next-generation models.