For most of the past two years, the loudest voices warning that artificial intelligence would hollow out white collar employment were not labour economists or unions. They were the chief executives building the technology. That makes the current reversal all the more striking. Within the same week in late May, both Sam Altman of OpenAI and Dario Amodei of Anthropic stepped back from the apocalyptic framing they had done so much to popularise.
From 50 percent to delighted to be wrong
The gap between the old message and the new one is wide. In a 2025 interview Amodei suggested AI could wipe out half of all entry level white collar jobs within five years and push unemployment towards 20 percent. Altman this spring struck a very different note, saying a global jobs apocalypse probably will not happen and that he was, in his words, delighted to be wrong, admitting he had expected far more impact on entry level work than has materialised. Amodei, in the same stretch, leaned into a narrative of augmentation rather than replacement.
Why the timing matters
The shift in tone has not gone unnoticed, in part because of what else is happening at both companies. OpenAI and Anthropic are reported to have filed confidential paperwork towards public listings, and analysts have been blunt about the incentive that creates. Institutional investors weighing a multi billion dollar offering tend to prefer a story of steady, productive adoption over one of mass social disruption. A founder talking down the upheaval his product might cause is, conveniently, also talking up its investability.
The data caught up with the rhetoric
Whatever the motive, the softer line is closer to what the statistics show. The Yale Budget Lab has found no significant change in the occupational mix or in unemployment duration for high exposure jobs since ChatGPT launched in late 2022. Employment across most knowledge work has held broadly stable while productivity has edged up, a pattern far removed from the cliff edge the early warnings implied. The real disruption, where it exists, has been narrow and concentrated rather than economy wide.
The lesson for workers and policymakers is to treat vendor forecasts in either direction with caution. The same executives who amplified the panic are now amplifying the reassurance, and both messages serve a commercial purpose. Plans for careers, training budgets and labour policy are better anchored to the labour data than to the mood of the people selling the models.
