Predictions on the record
Explore Sam Altman's AGI and superintelligence timeline: 12 sourced statements from Mar 3, 2015 to Oct 5, 2026. Compare what was said, when it was said, and any predicted arrival date.
This is a curated record, not a complete history or a scorecard. AGI and ASI labels group related language; the original wording and caveats remain below. How we document claims.
Prediction · ASI
No arrival date specified
“superhuman machine intelligence (SMI) is probably still decades away”
Altman argued for beginning safety and governance work despite a potentially long wait. His milestone was superhuman machine intelligence; the post did not announce an existing AGI.
Prediction · AGI
Expected 2027–2028
AGI could be reached in the next four or five years.
Source summary · not a verbatim quotation.
TIME reported his expectation in its CEO of the Year profile. The statement describes a forecast, not a released model.
Prediction · ASI
No arrival date specified
“It is possible that we will have superintelligence in a few thousand days”
Altman explicitly allowed that it could take longer. This forecast concerns superintelligence, a higher bar than many definitions of AGI.
Prediction · AGI
Expected by 2029
Within five years, people may look back on an AGI milestone that has already passed.
Source summary · not a verbatim quotation.
He expected rapid technical progress while the immediate effect on society could be smaller than people imagined.
Prediction · AGI
Expected 2025
“AGI.”
This was his answer when asked what he was most excited about in 2025. It was a brief expression of expectation, not a detailed technical prediction.
With a caveat · AGI
Arrival claim with a caveat
“We are now confident we know how to build AGI as we have traditionally understood it.”
He expressed confidence in a path to building AGI, rather than announcing that it had already been completed. The essay also anticipated agents joining the workforce in 2025.
With a caveat · AGI
Arrival claim with a caveat
Systems that start to point toward AGI are coming into view.
Source summary · not a verbatim quotation.
Altman described approaching a threshold, not having conclusively crossed it. His working definition involved handling complex problems at human level across many fields.
With a caveat · ASI
Arrival claim with a caveat
“Humanity is close to building digital superintelligence”
The essay described current systems as approaching AGI and helping accelerate AI research. It also acknowledged their remaining limitations and did not identify one model as conclusively AGI.
Prediction · ASI
No arrival date specified
“In ten more years, I believe we are almost certain to build superintelligence.”
The anniversary essay predicted a threshold beyond ordinary human-level AGI. It described a personal expectation rather than a demonstrated technical result.
With a caveat · AGI
Arrival claim with a caveat
AGI may already have passed as a recognizable milestone.
Source summary · not a verbatim quotation.
Altman floated treating AGI as crossed and focusing on superintelligence, while acknowledging disagreement over what AGI means. He did not establish a shared test or name a qualifying model.
With a caveat · AGI
Arrival claim with a caveat
“We basically have built AGI, or very close to it.”
Days after making the remark, he clarified that he meant it spiritually rather than literally and said additional breakthroughs were still required.
With a caveat · AGI
Arrival claim with a caveat
“I think a lot of people would say AGI happened or is close to happening.”
Asked about AI terminology in a POLITICO Decoded interview with Brendan Bordelon, Altman described what others might consider AGI's arrival while leaving open whether it was only near. He supplied no arrival deadline or model-specific claim. The wording comes from a third-party transcript; the original recording has not been independently verified.