Sam Altman has been one of the loudest voices warning about AI's impact on work. He said AI would "probably replace most of the jobs people do today," that customer support jobs are "totally, totally gone," and that the speed of change could be "potentially a little scary." This week, he posted that he is now "pretty sure" AI has been net job-creating so far, calling the result "not what I expected."
AnthropicCEO Dario Amodei made a similar journey in reverse. As recently as 2025, he was warning that AI could eliminate up to half of all entry-level white-collar jobs within five years and push unemployment to 10 to 20 percent. He has now repositioned that same technology as a productivity multiplier that expands what workers can do.
The data does support a calmer reading than either of them was giving a year ago. Yale's Budget Lab found no significant link between AI exposure and unemployment through mid-2025. A study covering 25,000 workers across 11 occupations in Denmark found no measurable effects on earnings or hours worked, even in firms actively encouraging AI use. A review of US labor market data found that where effects appear, they are mostly concentrated at the entry level of a few specific occupations, while senior employment stays stable.
There is a wrinkle, though. Tech layoffs through May 2026 have already passed 115,000 for the year. Companies from Amazon to Block have cited AI as the reason for cuts, but when New York state gave employers the option to officially cite "technological innovation" in legally required layoff notices, none of 160 companies filing notices checked that box. Researchers have a name for this pattern: "AI washing," using AI as a publicly acceptable cover story for cuts driven by overhiring during the pandemic, slower growth, or shareholder pressure. Nearly 60 percent of US hiring managers in a recent survey admitted they emphasize AI's role in reducing headcount because investors view it more favorably than admitting financial strain.
This is the moment to hold both things in mind. The sweeping job losses that Altman and Amodei were predicting have not materialized in the data, and that is genuinely good news. At the same time, both are now running companies preparing for stock market listings, with OpenAI targeting a valuation approaching one trillion dollars and Anthropic having already raised private funding at 965 billion dollars. A narrative about AI creating jobs is far easier to sell to public investors than one about mass unemployment.
For business operators, the practical picture has not changed much. AI is making certain tasks faster in specific roles, the productivity gains are real but uneven, and the labor market as a whole has not been reshaped yet. The honest answer is that nobody, including the people building this technology, knows the medium-term outcome with any confidence. What you can watch for in your own industry is whether companies citing AI efficiency in their public communications are actually deploying it at the operational level, or simply using the language to manage a stock price.
Altman's earlier warnings were not baseless; they reflected a genuine uncertainty about a fast-moving technology. His current reassurance deserves the same scrutiny. Both positions were made in public, with an audience of investors paying close attention.