Sam Altman: 'The World Is Not Prepared' for What's Coming in AI

Sam Altman sat down with Anant Goenka at Express Adda in India today and spent nearly an hour walking through his current thinking on AI. The interview covered a lot of ground: India’s AI ecosystem, job disruption, China, a not-very-explained awkward moment with Dario Amodei, and a rapid fire section that got genuinely interesting toward the end. But one answer stopped the conversation cold.

Asked about the resignation of Anthropic’s safety chief, Altman said this:

“The part of it I agree with is the inside view at the companies of looking at what’s going to happen. The world is not prepared. We’re going to have extremely capable models soon. It’s going to be a faster takeoff than I originally thought. And that is stressful and anxiety-inducing.”

That quote is from the CEO of OpenAI. He said it in the middle of a polished keynote-style interview, without hesitation. Take it seriously.

What He Actually Said About the Timeline

The interview opened with Altman painting a timeline of AI capability that’s worth sitting with. A year ago, he reminded the audience, AI could do “okay” at high school math. By last summer, it was competing at the hardest international math competitions. Last week, a model solved seven out of ten unsolved research-level math problems put out by professional mathematicians.

“AI has gone from doing okay at high school math to being able to do new research-level mathematics, figure out new knowledge,” he said. “This is an amazing change in a year.”

He also said that AGI, which he’s historically placed a few years out, now “feels pretty close.” And then he went a step further: “Given what I now expect to be a faster takeoff, I think super intelligence is not that far off.”

He said both things calmly, in the same breath as talking about India’s startup ecosystem.

The Safety Chief Moment

The quote that’s going to circulate was technically in response to a question about Anthropic’s safety chief resigning. The host framed it loosely as someone who’d decided AI safety work was hopeless and wanted to go write poetry. Altman didn’t dispute the characterization of the anxiety behind it.

“The part of it I agree with” is a careful phrasing. He’s not endorsing resignation. He’s endorsing the underlying feeling: that the people closest to these systems see something the public doesn’t, and that something is accelerating. When you run OpenAI, saying “I agree that the inside view is stressful and anxiety-inducing” is a meaningful thing to put on record.

Jobs, India, and the Honest Version

A significant chunk of the interview was about India, where Altman noted that Codex, OpenAI’s coding tool, is seeing its fastest global growth. He was warm about the builder energy he’s seeing there, particularly after visiting IIT Delhi that morning.

On the IT services sector that makes up about 8% of India’s GDP, he didn’t soften it: “I do think that is going to change a lot and there’s going to be a big impact there. And I think it is never helpful to pretend otherwise.”

His follow-up was more optimistic. He’s not a “jobs doomer,” he said. Every technological revolution has triggered the same panic and found new things to do on the other side. He made the fair point that no one in the industrial revolution predicted YouTube influencers or AI CEOs. The specific jobs that emerge are almost impossible to predict; what matters is adaptability and fluency with new tools.

He also shared a useful contrast on creative work: when AI image generation appeared, everyone predicted the end of graphic artists. What actually happened was that commercial commodity art (birthday invitations, stock illustrations) got crushed by free AI output, but original human-made fine art continued to appreciate. Price of AI art: zero. Price of human art: going up.

The Corporate Adoption Mistake

One of the sharper moments in the rapid fire section: Altman described a meeting from the day before with a large company that had mapped out its AI strategy as follows: spend 2026 strategizing, 2027 getting the organization ready, and 2028 deploying.

“That may work for other kinds of technology,” he said. “Apparently, if you do like a giant ERP migration, that’s the kind of timeline it takes. Doing that for AI will be a catastrophic mistake.”

The companies losing the most ground right now are the ones treating this like a managed IT transformation with a three-year runway.

China, Democracy, and Who Should Lead

Altman’s China take was measured but clear. China is “very ahead” in some areas: physical robotics, energy infrastructure, electric motors. In other areas, the US still leads. He pushed back on the binary framing, noting that it’s hard to be ahead or behind on everything simultaneously.

His consistent position: no single country, company, or AI system should be in charge of superintelligence. “I think the world is at its best when power is widely distributed.”

When asked to give the same one-sentence advice to Xi Jinping, Narendra Modi, Putin, and Trump, he gave the same answer to all four: “You got to democratize this technology and put it in the hands of people.”

He admitted not all of them would listen.

The Awkward Dario Moment

The host referenced an “awkward moment” on stage the previous day that had apparently produced a round of memes. It involved Dario Amodei, the former OpenAI safety head who left to found Anthropic, which now directly competes with OpenAI and has staked out a notably more cautious public tone on AI risk.

Altman’s response: “I don’t really have that much more to add.”

He did say, in the broader context of competition, that despite commercial rivalry, all the serious AI labs are “very committed to getting safety and alignment right and willing to cooperate there.” Whether that’s reassuring depends on your priors.

Elon, Equity, and the Stuff He’s Tired Of

On Elon Musk reconciliation versus TSMC losing its semiconductor monopoly, Altman said Musk and he becoming friends again is “less likely.” Then he added: “I feel like I have more control over that one.”

On his decision to take no equity in OpenAI, he called it “truly one of the dumbest things” he’s done. Asked if he’d take equity now if someone figured out a way to offer it: “I feel so tired of the whole conversation and so trapped I’m not sure what to do.”

The one thing he’d never ask ChatGPT: how to be happy. “I would rather ask a wise person.”

The Thing That Actually Matters

The full interview is worth watching. Altman is good at this format. He’s specific, often funny, and doesn’t usually retreat to pure boilerplate.

But the “world is not prepared” comment is the one. Not because it’s a new sentiment — Altman has gestured at existential risk before — but because of the context it came in. He wasn’t being pressed on AI safety. He was responding to a question about a safety researcher who burned out. His answer was, essentially: I understand why they burned out. I see what they saw.

“Faster takeoff than I originally thought” from the founder and CEO of OpenAI, in a casual answer to a side question, is a different kind of statement than the same words in an AI safety paper or a congressional hearing. It’s what you say when you’re not performing concern. It’s what you say when you mean it.

Sources

• Sam Altman at Express Adda — YouTube — full interview with Anant Goenka, February 20, 2026
• @kimmonismus on X — viral clip of the key quote