AI Was Going to Kill Millions of Jobs. A Year Later, the Numbers Tell a Different Story.

Key points

• In May 2025, Anthropic CEO Dario Amodei predicted AI would eliminate half of all entry-level white-collar jobs.
• One month later, OpenAI CEO Sam Altman said certain job categories would disappear entirely.
• Companies began citing AI as a reason for layoffs in 2025, prompting worker organizing and student anxiety.
• Despite those predictions, no mass wave of AI-caused unemployment has appeared in the data so far.
• Economists say the picture is still shifting, and more change is expected.

The carnage was supposed to be obvious by now.

In May 2025, Dario Amodei, chief executive of Anthropic (the company behind the Claude AI assistant), told an audience that AI would wipe out half of all entry-level white-collar jobs. White-collar work means desk jobs: accountants, paralegals, customer-service agents, junior analysts. Half. Gone.

A month later, Sam Altman, the chief executive of OpenAI (the company behind ChatGPT), went further. Entire job categories, he said, would simply stop existing.

So what actually happened?

The predicted wave never arrived, at least not in the employment statistics. Job losses tied directly to AI have not shown up at the scale either executive suggested, according to reporting first flagged by The Guardian AI.

That does not mean nothing changed. Companies did start citing AI in layoff announcements. Workers organised in response. Students began rethinking degree choices and career paths.

But "beginning to change" and "mass carnage" are very different things. The labour market has not collapsed.

Why the gap between prediction and reality?

Economists point to a familiar pattern: transformative technologies tend to shift work before they eliminate it. A spreadsheet did not fire accountants overnight; it changed what accountants spent their days doing. The same logic may apply here.

Large language models, the technology behind tools like ChatGPT and Claude that can read, write and answer questions, are getting faster and cheaper. But deploying them inside real businesses takes time. Legal reviews, retraining, liability questions, union agreements and plain organisational inertia all slow the process down.

That lag is important. It means the disruption economists expect may still be coming, just on a longer timeline than a May 2025 headline implied.

What does this mean for ordinary workers?

If you work a desk job involving writing, research, data entry or customer communication, the honest answer is: stay alert, not panicked.

The evidence so far suggests AI is changing how tasks get done inside those roles faster than it is eliminating the roles themselves. Employers are beginning to expect workers to use AI tools the way an earlier generation was expected to learn email or Excel.

That is not a guarantee of safety forever. Economists do expect more change ahead. But the timeline appears to be years, not months.

Common questions

Are AI-related layoffs actually happening?

Yes, some companies have cited AI when announcing job cuts, particularly in customer support and entry-level content roles. But economy-wide data does not yet show AI as a leading cause of unemployment.

Should students avoid certain degrees because of AI?

Not based on the current evidence. Careers involving judgment, relationships and physical presence remain largely unaffected so far. The more useful question is which skills within any field AI tools can strengthen, not which fields to abandon.

Will this change?

Economists broadly expect AI to reshape more jobs as the technology matures and spreads. The current picture is a pause, not a verdict.