They Predicted Mass Job Losses. Now They Say It Wasn't So Bad.

In 2025, Dario Amodei, the chief executive of Anthropic, put a number on what his own technology would do to the workforce.

roughly 50% of entry-level white-collar jobs within 1–5 years, and that unemployment could spike to 10–20% within one to five years. — Dario Amodei

Fourteen months later, in May 2026, the same man describing the same technology reached the opposite verdict.

The narrative that connects AI to job loss, for many of the CEOs that are doing it -- it is just too lazy — Jensen Huang

The technology did not change its mind. The people selling it did. Sam Altman, who had once predicted a "punctuated equilibria moment" in which customer-support workers would lose their jobs, told interviewers this spring that he had expected more entry-level white-collar jobs to be eliminated by now than has actually happened ^[1][2]. Jensen Huang called the AI-layoff narrative "lazy" and "irresponsible." The reversal was nearly total, and it arrived just as OpenAI and Anthropic were preparing to go public at valuations near a trillion dollars each ^[2]. While the executives were revising their forecasts, two independent datasets were landing that told a different story. Stanford researchers, working from ADP payroll and Bureau of Labor Statistics data through June 2026, found no economy-wide job losses — but they did find the entry-level employment gap widening from 15% to 19% in a single year, driven not by layoffs but by companies simply hiring fewer young workers ^[3]. California's new statewide tracker, launched in June, found no large-scale AI-related layoffs and, in the same report, a sustained rise in unemployment claims since 2022 among college-educated workers in high-exposure occupations ^[4]. The two findings are not in conflict. The top line is fine; the bottom rung is disappearing. The walk-back was not a passive change of heart. In April, OpenAI acquired a podcast network, and the company was explicit about why ^[5].

The biggest reason is simply that the rate of progress is continuing to accelerate, and we believe we are very close now. — Sam Altman

Palantir launched its own campaign, the "AI Optimism Project," to counter what its chief executive called a lack of understanding that fuels populism ^[6]. Both OpenAI and Anthropic also began paying consulting firms to work on their workforce messaging ^[5].

I do suspect that we're gonna have to make changes to how we tax like in a world where AI is doing Most of the intellectual work in the world or at least of the work of today. — Sam Altman

None of this is hidden. It is documented, budgeted, and public — which is what makes it more striking than any conspiracy would be. The policy response splits along the same fault line. In January, China's Ministry of Human Resources announced a framework with particular priority on supporting university graduates and young jobseekers — the exact demographic the Stanford data shows being hollowed out ^[7]. The U.S. administration's answer was different. White House AI czar David Sacks reached for the buggy whip.

It’s really easy to see, in a big breakthrough, that the horse-and-buggy guys are going to go out of business. — David Sacks

The administration's AI Action Plan acknowledges widespread job displacement as a risk, then proposes to manage it by deploying more AI agents — using AI to supervise other AI systems for compliance ^[8]. The remedy for displacement is more displacement. To be fair to the optimists, the aggregate numbers they cite are real. Stanford found no economy-wide job losses, and new AI-related roles are being created by the million ^[9]. The harm is specific and demographic — a hollowing-out of the entry-level rung for college-educated workers — which is exactly why pointing at top-line employment is a sleight of hand rather than a lie. The people being told it has not been as bad as expected are the ones for whom it has been. The most telling voice came from inside the apparatus. Christopher Olah, a co-founder of Anthropic, stood at the Vatican and contradicted his own chief executive.

We, as the producers of this technology, have a duty and an obligation to be honest about what is coming. — Dario Amodei

A company preparing to go public does not usually send its co-founder out to say the thing its communications strategy is working to suppress. That Olah said it anyway is the clearest measure of how wide the gap has grown between what the industry's leaders forecast in private and what they are willing to say in public.