Altman Reportedly Outlined Three Paths to Safe AI, Including a Single Firm With Power to Shut Down Rivals
OpenAI CEO Sam Altman reportedly presented three scenarios for governing advanced AI in an internal talk: an international watchdog modeled on the IAEA, a merger of leading labs, or one dominant company able to disable competitors' computing power. He called the last option «terrible».
OpenAI chief executive Sam Altman has reportedly laid out three possible endgames for keeping advanced artificial intelligence safe, including a scenario in which a single company becomes powerful enough to shut down every other firm's computing infrastructure. The options were presented in an internal talk and range from a multinational regulator to a merger of the leading AI laboratories.
According to the account, the first path envisions an international watchdog modeled on the International Atomic Energy Agency, the United Nations body that monitors nuclear materials and facilities. Such a body would oversee the development of frontier AI systems in the way the IAEA inspects civilian nuclear programs, giving governments a shared mechanism for verification and restraint.
The second option would see the major AI labs merge into a single entity, consolidating research and safety decisions under one roof rather than competing across separate companies. The third and most drastic scenario describes a single firm so dominant that it could disable the computing power of everyone else — an outcome Altman reportedly described as «terrible».
The detail that drew the most attention was the image of one company melting or shutting down other firms' GPUs, the specialized chips that underpin modern AI training and deployment. The phrase captures a concentration of power that would leave a single actor holding a de facto switch over the industry's hardware.
The reported presentation comes as governments and international bodies debate how to govern increasingly capable AI systems. Proposals have ranged from licensing regimes and safety testing to calls for a global agency that would coordinate oversight much as nuclear safeguards are coordinated today. The IAEA comparison has become a recurring reference point in those discussions because it offers a template for international inspection and shared rules.
For OpenAI, the disclosure offers a glimpse of how its leadership frames the long-term risks of the technology it is building. The company has positioned itself as a proponent of safety measures and has argued that frontier models require oversight beyond what any single firm can provide. At the same time, the merger and dominance scenarios raise questions about competition and the concentration of control over critical computing resources.
The reported options also reflect a broader tension in the AI industry: the same companies developing the most powerful systems are also among those calling for rules that would govern them. Whether oversight should come from a multinational body, from consolidation among labs, or from market dominance remains unresolved, and the three scenarios illustrate how differently the endgame could look.
No formal proposal appears to have followed the internal presentation, and the scenarios were described as paths rather than a policy plan. Still, the account adds to a growing public record of how senior AI executives think about the risks and governance of systems that are expected to grow more capable in the coming years.
4
