A year ago, the two most watched executives in AI put a number on fear. Dario Amodei of Anthropic said AI could erase half of entry level white collar jobs within five years. Sam Altman of OpenAI said whole categories of work would vanish. Those lines traveled everywhere, from boardrooms to graduation speeches.
The numbers since then do not back up the prediction. A Stanford analysis found that unemployment for workers most exposed to AI rose by well under one percentage point since 2022, and by less than it rose for workers barely touched by AI at all. Employers have cited AI in roughly 55,000 job cuts this year, according to Challenger, Gray and Christmas, against more than 1 million total layoffs. AI shows up in the layoff data, but it is nowhere near the main cause.
What makes this more interesting is that the men who made the boldest predictions are now walking them back. Altman recently said he was wrong about how much entry-level work would disappear by now, and that he was glad to be wrong. Amodei has shifted his language too, describing AI as something that lets workers do more rather than something that replaces them outright.
The timing is not a coincidence. Both companies are heading toward stock listings that could value them near or above a trillion dollars, with Anthropic targeting a listing later this year and OpenAI possibly following in 2027. A story about mass job destruction is a hard sell to public market investors who need to believe the technology grows the economy rather than hollowing it out. Warnings that once generated headlines are now a liability on the road to an IPO.
None of this means the labor market is calm. It means the damage is showing up differently than advertised. Companies are not firing waves of workers because of AI. They are quietly refusing to backfill roles, raising the bar for new hires, and asking existing staff to do more with the same headcount. A recruiting survey found that most employers now treat AI skill as a strong advantage or a flat requirement, and half expect new hires to already be capable users on their first day.
There is also a longer term shift worth watching. Research from MIT puts the share of the US workforce already treated as disposable, meaning contractors, freelancers, and workers without real job security, at around one third. Employers who are unsure exactly which skills they need are more likely to hire this way while they figure it out, and AI adds another reason for that uncertainty.
For a business leader, the lesson is not to relax because the mass layoff never came. It is to notice that hiring bars are rising quietly, job security is thinning at the edges, and the loudest voices predicting disaster were also the ones with the biggest financial reason to later say they were wrong. Judge the technology by what companies are actually doing with headcount, not by what its builders say in interviews.