Sam Altman

CEO and co-founder of OpenAI

OpenAI

Lab leaders

Altman runs the company behind ChatGPT and is racing to build superintelligence while saying it is the most dangerous technology humanity has made. He avoids giving a probability of catastrophe, has told Congress the technology 'can go quite wrong', and now says the 'takeoff has started', with fully automated AI researchers targeted for 2028.

Declines to give a number

No numeric p(doom) on record. In 2015 he wrote: "Development of superhuman machine intelligence (SMI) is probably the greatest threat to the continued existence of humanity." In Senate testimony (May 16, 2023): "if this technology goes wrong, it can go quite wrong."
Source

On timing

"We are now confident we know how to build AGI as we have traditionally understood it" (Reflections, January 2025). "We are past the event horizon; the takeoff has started... 2026 will likely see the arrival of systems that can figure out novel insights. 2027 may see the arrival of robots that can do tasks in the real world" (The Gentle Singularity, June 2025). Internal goals: "an automated AI research intern by September of 2026... and a true automated AI researcher by March of 2028" (X, October 29, 2025).

Source

What they call for

  • A new federal agency that licenses frontier scale efforts (2023 testimony)
  • Solve the alignment problem
  • Make superintelligence 'cheap, widely available, and not too concentrated'
  • Iterative deployment

In their own words

  • "I would form a new agency that licenses any effort above a certain scale of capabilities and can take that license away" (Senate Judiciary subcommittee).

    16 May 2023

    Source

  • "We are past the event horizon; the takeoff has started." Priorities: "Solve the alignment problem... Then focus on making superintelligence cheap, widely available, and not too concentrated with any person, company, or country."

    10 June 2025

    Source

  • "We have set internal goals of having an automated AI research intern by September of 2026 running on hundreds of thousands of GPUs, and a true automated AI researcher by March of 2028. We may totally fail at this goal."

    29 October 2025

    Source

Signed

  • CAIS extinction statement 2023

Last verified 1 September 2026

Report a correction

Back to all voices