The man who built ChatGPT just walked back one of his most quoted predictions. Make of that what you will.
At a Commonwealth Bank conference in Sydney on May 26, OpenAI CEO Sam Altman told the audience he’d been “pretty wrong” about AI’s impact on jobs. “I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened,” he said. “I’m delighted to be wrong.”
That’s a striking reversal from the guy who, not long ago, was warning that AI would handle 30–40% of all tasks in the economy. Now he’s standing on a stage in Australia saying the jobs apocalypse he helped predict isn’t happening.
What he actually thinks AI won’t replace
According to Altman, the roles hardest for AI to displace are those built around human emotion, trust, and personal connection — healthcare, education, consulting, mentorship, customer-facing work, and any job where the relationship itself is the product. These aren’t glamorous categories, but they’re also not going anywhere.
The logic holds up. An AI can summarize a patient’s chart. It can’t hold someone’s hand in a hospital room. It can generate a lesson plan. It can’t notice that a kid in the back row stopped raising their hand three weeks ago.
The data is messier than the headline
Here’s where it gets complicated. The Yale Budget Lab found no meaningful change in unemployment through March 2026 for workers in jobs with high AI exposure — which supports Altman’s softer take. But zoom in a layer and the picture shifts. Over 113,000 tech workers have been laid off across 179 companies since January 2026 — a pace 33% higher than the same period last year, with AI cited as a driver at Meta, Amazon, and Snap among others.
So the macro numbers look stable while specific sectors are getting hit hard. Both things are true at once.
The IPO problem nobody wants to say out loud
OpenAI reportedly filed IPO paperwork confidentially in May 2026, with a target listing in September or October. A company valued at over $1 trillion cannot go public while its CEO is publicly predicting mass unemployment driven by its own products. The timing of Altman’s reversal — landing within weeks of that filing — has not gone unnoticed.
There’s also a harder-to-ignore contradiction: earlier in 2026, OpenAI published a policy paper calling for taxes on automated labor, a national public wealth fund, and pilots of a 32-hour working week — a document that explicitly assumes significant labor disruption is coming. That paper and Altman’s Sydney remarks don’t easily coexist.
Meanwhile, Anthropic CEO Dario Amodei — who once warned AI could dissolve up to 50% of entry-level white-collar jobs within five years — has also quietly shifted to emphasizing AI as augmentation rather than replacement, right as Anthropic eyes its own fundraising milestone.
Maybe both CEOs genuinely updated their views based on evidence. Or maybe there’s a financial calendar that makes optimism more convenient right now. Either way, the workers whose jobs actually changed this year probably aren’t delighted.