The AI jobs apocalypse got postponed.
A year ago, the two biggest names in AI told us our jobs were in trouble. OpenAI's Sam Altman warned that entire categories of work would vanish. Anthropic's Dario Amodei even put a number on it: roughly half of entry-level white-collar jobs gone within five years, unemployment climbing toward 10 or 20 percent.
This month, both of them walked it back. Altman now says he's "delighted to be wrong." Both companies, as it happens, are weeks from asking public investors for a mountain of money.
Here's the short version of the about-face. In 2025, the message was urgent: a technology so powerful it would reshape work within a few years, and everyone needed to brace for it. In May 2026, Altman told a banking conference he'd expected far more damage to entry-level jobs by now than has actually shown up. Amodei softened too. He now talks about automation less as a job-killer and more as a way to get more done with the same people.
What shifted over the past year was the evidence, not so much the technology. The Yale Budget Lab has tracked AI's effect on US employment since ChatGPT launched, and through March 2026 it found no sign that AI is changing which jobs people hold or how long they stay unemployed. This is true even in the roles most exposed to it. The disruption that was supposedly already underway hasn't happened.
The timing of these comments could be suspect. Both OpenAI and Anthropic are preparing to go public this year at valuations in the hundreds of billions. Predicting mass unemployment would be a strange thing to pitch to investors who want growth, and to a public that already feels uneasy about AI. One policy researcher quoted by Fortune put it plainly: it's hard to tell whether the CEOs changed their actual forecasts or just changed the story they tell. The layoffs themselves are real (tech cuts passed 115,000 this year, with Meta, Intuit and others naming AI as a reason), but "we're using AI now" has become a convenient label for cuts companies wanted to make anyway.
Here's what this means
I've said before that the useful posture toward AI is boring: it's a tool, you should learn to use it, and you should ignore most of what the people selling it tell you about the future. This story is that boring advice proving itself out. The same executives who wrote the doom narrative are revising it every few months, right in step with their fundraising.
If you run a small business or a nonprofit, the practical reality is the same, despite the headlines. AI can save you time on tasks. It's not about to run your operation or replace your people. Yet.
What to do
- Don't make staffing or career moves off a prediction. The forecasts have a track record now, and it's poor in both directions.
- Judge AI by what it does on your actual work. Try it on a task you do every week. That tells you more than any keynote from a CEO.
- Watch the layoff language. When a company blames AI for cuts, ask whether the math worked without it. It probably will.
- Keep learning the tools anyway. An overhyped apocalypse doesn't make the skills worthless.
The walk-back isn't a confession that AI won't change work. Some jobs will keep shifting, and the slow, undramatic version of that is probably the real story. What got cancelled was the timeline, the one that sold well in 2025 and sells differently now that there's stock to move and money to be made. Read the next big prediction with that in mind.
The data: Yale Budget Lab, AI labor market tracker
