OpenAI CEO Sam Altman has outlined three possible scenarios for the future of artificial intelligence, ranging from an international watchdog overseeing advanced AI systems to a unified research organisation bringing leading AI labs under one roof. The third possibility involves a single company gaining overwhelming computational power and eliminating its rivals, a scenario Altman reportedly described as “terrible”.According to The Latin Times, which reported on October 19, 2026, citing an internal presentation to OpenAI employees, Altman discussed these possibilities as competing approaches to governing the development of increasingly powerful AI systems. The discussion comes as technology companies and policymakers grapple with questions around AI safety, accountability and the risks of an uncontrolled race towards superintelligence.
An international watchdog for AI safety
Altman’s preferred approach, as outlined in the report, involves creating an international oversight body modelled on the International Atomic Energy Agency (IAEA), which monitors the peaceful use of nuclear technology.Such a body could independently inspect advanced AI systems, establish safety standards and require audits before powerful models are deployed. It could also monitor the computing infrastructure and energy resources used to develop frontier AI.The proposal builds on an idea Altman has previously advocated. In May 2023, he and fellow OpenAI leaders called for an international institution resembling the IAEA to help oversee the development of superintelligent systems.However, implementing such a framework would require cooperation between countries competing for technological leadership, particularly the United States and China.
The ‘Demis Plan’: Bringing AI labs together
The second scenario, which Altman reportedly called the “Demis Plan”, draws inspiration from CERN, the European organisation that brings countries together for scientific research.Named after Google DeepMind CEO Demis Hassabis, the proposal envisages leading AI laboratories consolidating their research efforts into a single organisation. The reasoning is that reducing competition between developers could ease pressure to release increasingly powerful systems before their safety has been adequately assessed.However, such consolidation could create its own challenges. Bringing the leading AI developers under one institutional structure could concentrate technological power and raise questions about competition, accountability and control over advanced AI research.
Why Altman rejects a compute monopoly
The third scenario describes a company accumulating enough computing resources and technological advantage to disable competing AI systems globally.The Latin Times reported that Altman referred to this possibility as the “Eliezer Plan”, associating it with AI safety theorist Eliezer Yudkowsky’s concerns about recursive self-improvement and the possibility of advanced AI systems pulling far ahead of human control.Altman reportedly called the scenario “terrible”, distinguishing it from the other two approaches to managing the future of AI.
AI industry moves towards self-regulation
Alongside these longer-term possibilities, major AI developers are reportedly exploring a more immediate regulatory mechanism.According to The Latin Times, Google, OpenAI and Anthropic are working towards establishing the Standards Authority for Frontier AI (SAFA), a proposed industry-led body modelled on the Financial Industry Regulatory Authority (FINRA).The initiative could focus on pre-release model testing, safety incident reporting and the accreditation of independent auditors. A launch is being targeted for late 2026 or early 2027, the report said.The proposed body would face questions about enforcement, independence and whether voluntary industry standards can adequately address risks posed by increasingly capable AI systems.Ultimately, the three scenarios highlight a central challenge for the AI sector: how to ensure that technological progress does not outpace the safeguards needed to manage it. Whether through international oversight, institutional cooperation or industry-led regulation, the debate over who controls advanced AI is becoming increasingly difficult to separate from the technology’s future.
