The OpenAI chief, speaking in the Asia-Pacific, walked back the more dramatic predictions of broad employment collapse. The data, so far, agrees with him.

OpenAI chief executive Sam Altman said on Tuesday that artificial intelligence was unlikely to trigger the broad employment collapse that has come to be known, in the industry’s own shorthand, as the jobs apocalypse, even as he conceded that specific categories of work, including customer support, will largely disappear.

The remarks, reported by Reuters from an Asia-Pacific appearance, mark the latest step in a noticeable softening of tone from the executives who, until recently, were the loudest voices warning about AI-driven labour disruption.

Altman himself has spent much of the past year describing customer service jobs as “totally, totally gone” in the near future, and saying that traditional work skills now have a two-to-three year half-life.

The newer framing, repeated across appearances in India, Japan, and South Korea over recent months, draws a different line: significant churn within sectors, yes; an economy-wide collapse in headcount, no. The shift coincides with the absence, so far, of the kind of macro signal that an actual jobs apocalypse would generate.

The Yale Budget Lab, which has been tracking AI’s effect on US labour markets since the release of ChatGPT, has found no meaningful change in occupational mix or unemployment durations through March 2026 for workers in jobs with high AI exposure.

Anthropic’s own usage data, incorporated in the lab’s February update, did not move the picture. The Brookings Institution reached a similar conclusion earlier this year: no apocalypse, at least not yet.That “not yet” is the part Altman has spent more time on.

At the India AI Impact Summit in February, he told CNBC-TV18 that some companies were engaging in “AI washing,” blaming layoffs on AI that they would have carried out anyway, while real displacement was nonetheless beginning to show up in particular roles.

He has been more direct about specific categories: customer service work done over phone or computer, in his view, will be replaced and better performed by AI within the next few years.

Coding has been reshaped already, with engineers spending less time writing code and more on architecture, system design, and reviewing AI-generated work.

The softer macro framing is not entirely in tension with OpenAI’s own policy work. The company published a 13-page policy document earlier in 2026 calling for taxes on automated labour, a national public wealth fund partly seeded by AI companies, and pilots of a 32-hour working week. The document presumes significant labour-market disruption is coming.

Altman’s May remarks are best read as a calibration of the timing and shape of that disruption, not a denial of it: less a single rupture, more a long rolling reshuffle in which some categories vanish, others change beyond recognition, and the headline employment number does not necessarily move much.

Altman’s recent travel itinerary has tracked the audience for that message. He visited Tokyo in the spring to meet SoftBank chief executive Masayoshi Son and Japanese prime minister Shigeru Ishiba, and was in Seoul shortly afterwards for a developer event hosted by OpenAI. His Asia-Pacific appearances have, as a rule, leaned harder than his US ones on the “new jobs will emerge” framing.

The Yale Budget Lab’s next data update is due in the coming weeks. Until then, the headline labour numbers and Altman’s framing of them sit in roughly the same place: stable, for now.