OpenAI's Altman Outlines Three AI Safety Paths, Says Decision Needed Within a Year

OpenAI CEO Sam Altman has described three possible paths for AI safety development to employees, saying the company needs to decide which one to pursue within the next year, according to The Information.

The three paths include establishing an international regulatory body similar to the International Atomic Energy Agency (IAEA), merging major AI labs into a single entity modeled on CERN, and allowing one company to monopolize computing power—thereby cutting off rivals' access to the chips needed to train advanced models.

Altman reportedly said the third path would be "clearly terrible" and that he included it only for completeness. He also stressed that humans should always remain in control of the technology and that industry power should not become overly concentrated.

Altman has long favored an IAEA-style oversight body for advanced AI, a concept he and OpenAI co-founders discussed publicly as early as 2023. The report noted that the first two paths would require extensive coordination among companies or governments, and that OpenAI currently "cannot afford to make concessions in the competition."