Ilya Sutskever, his brain behind Chatgpt, spoke of an astonishing plan in front of his colleagues from Openai. An “apocalypse bunker” to house the researchers of the company at the dawn of an artificial intelligence which would go beyond humanity.
Ilya Sutskever, one of the brains behind Chatgpt, seriously plans to build a bunker to protect his colleagues from Openai. His fear: than an artificial intelligence more powerful than humans triggers chaos and violence.
The idea does not come from an anticipation novel, but from the heart of Silicon Valley. In the summer of 2023, during an internal meeting at Openai, Ilya Sutskever, co -founder and then scientific director of the company, launched to his colleagues a sentence that leaves them speechless: “Once we are all in the bunker …” In front of the general surprise, he specifies his thought: before revealing an AG, a general artificial intelligence surpassing the human, it would be necessary to prepare a hiding over for the main researchers The outside world would rock in panic or violence.
The episode is reported in “Empire of AI: Dreams and Nightmares in Sam Altman’s Openai”, the book by Karen Hao, published in May 2025. And the proposal is not limited to a simple formula launched on the fly. Several witnesses interviewed by Karen Hao confirm that the term regularly returns to internal discussions, proof of an obsession shared by certain key members of the team. In the mind of Ilya Sutskever, AGA, this general artificial intelligence that the scientific community dreams and fears both, would not only upset technology: it would be likely to provoke an unprecedented political and social shock wave.
If Openai managed to create the first digital entity truly superior to humans, states around the world could embark on fierce, even violent competition, to take control, estimates the sources of the journalist. Faced with this prospect, a simple laboratory or secure offices would not be enough. It would take a fortress, a shelter capable of resisting all the crises.
The bunker imagined by Ilya Sutskevern is not described in his smallest detail, but the logic underlying the project is clear. It would be a refuge reserved for the main OpenAi brains: a secret, protected place, where researchers could come together from any pressure or any external threat, the time that the world adapts to the new reality created by AG. ILYA STSKEVER insists: participation would be voluntary, but it considers essential to predict this outcome of aid.
According to Karen Hao’s investigation, some collaborators go so far as to speak of “Rapture”, a religious term designating the sudden disappearance of an elite, to qualify this scenario where some elected officials would withdraw from the world from the threshold of a technological revolution. Officially, the bunker project will never be officially launched, but it leaves a lasting trace in the internal culture of Openai.
