Sam Altman has put an unusual idea on the table: the world’s biggest AI labs could one day combine into a single, cooperative organization in the name of safety.
According to a report from The Information on October 19, 2026, the OpenAI CEO walked employees through three possible futures for keeping advanced AI in check. The merger concept was one of them. It was not, by the outlet’s account, his favorite.
Three roads, one preferred route
Altman framed the internal discussion around the need to act before problems arrive, not after. He then sketched out three scenarios, each with a very different picture of who holds the keys to powerful AI.
Scenario one is an international oversight body modeled on the International Atomic Energy Agency, or IAEA. Sources cited in the reporting indicate this is the option Altman leaned toward.
Scenario two is what Altman reportedly called the “Demis plan.” Under this approach, leading AI organizations would fold into one collaborative entity. The model here is CERN, the European physics research center where member nations pool resources on shared science. Applied to AI, the merged group would develop the technology on behalf of participating countries rather than for any single company’s shareholders.
The idea traces back to a concept Demis Hassabis put forward for a CERN-like AI institution nearly a decade ago.
Scenario three is dubbed the “Eliezer plan” after AI safety researcher Eliezer Yudkowsky. It imagines one dominant firm using advanced AI to wipe out its competition. Altman reportedly described this outcome as harmful and included it only so the discussion would cover every possibility.
Why rivals may not hold hands
OpenAI and Anthropic are both pursuing IPOs. Companies preparing to go public tend to sharpen their competitive edge, not hand it to a consortium.
Public debate over safety frameworks has been running for weeks, with OpenAI, Anthropic and Google DeepMind all part of it. Behind closed doors, the labs have also been holding private talks about safety standards. Altman’s comments landed in the middle of that back-and-forth.
Washington enters the chat
The Trump administration is preparing its first regulatory framework for AI development. Any federal rulebook would shape which of Altman’s three scenarios is even realistic. An IAEA-style international body would require governments to agree on shared oversight across borders. A CERN-style merger would raise questions for regulators, as folding the leading labs together would concentrate a great deal of capability in one place.
What this means
For investors watching the IPO pipeline, increased oversight could bring heightened scrutiny to AI companies preparing to list. If the merger route ever gained traction, it could lead to consolidation across the sector.
The key things to watch are the shape of the federal framework, whether the private talks among OpenAI, Anthropic and Google DeepMind produce any public commitments, and how the IPO timelines for OpenAI and Anthropic evolve.
Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.