The Canaries Are Already Singing

13 percent. That's how much employment dropped among workers aged 22 to 25 since late 2022, according to Stanford researchers who analyzed millions of payroll records.

They called it "The Canaries Paper"—a reference to the birds miners once used to detect poison gas. Like those canaries, young workers might be the first to signal danger in the AI revolution.

The Silence After the Storm

Something strange happened in early 2025. After months of CEOs openly discussing AI's impact on their workforces, the pronouncements stopped. Eerily. All at once.

In May 2025, Anthropic CEO Dario Amodei had warned that AI could drive unemployment up 10 to 20 percent in the next one to five years. Ford's Jim Farley estimated it would eliminate "literally half of all white-collar workers" in a decade. OpenAI's Sam Altman revealed his "little group chat with tech-CEO friends" had a bet about when a billion-dollar company would be staffed by just one person.

Then silence. Amazon, Meta, Walmart, JPMorgan Chase—all declined interview requests. Even the Business Roundtable, which exists to speak for America's most powerful CEOs, had "nothing to say."

The simple explanation: 280,590 public-relations specialists now employed in America saw which way the wind was blowing. AI is unpopular. CEOs talking about job cuts are even less popular. So they shut up.

The Economists' Dilemma

"Numerically speaking, nothing indicates that AI has had an impact on people's jobs," says Austan Goolsbee, president of the Federal Reserve Bank of Chicago. "It's just too early."

Economists are constrained by numbers, and the numbers tell a familiar story. ATMs led to more bank tellers. Excel created more accounting jobs. Photoshop increased demand for graphic designers. Technology destroys old tasks but creates new, better-paying work.

But Anton Korinek at the University of Virginia thinks his colleagues are misreading the technology. "We can't quite conceptualize having very smart machines," he told me. "Machines have always been dumb, and that's why we don't trust them and it's always taken time to roll them out. But if they're smarter than us, in many ways they can roll themselves out."

The Speed Problem

This is where AI differs from previous disruptions. The China shock of 2001-2007 eliminated 2 million manufacturing jobs over six years—devastating for affected communities, but slow enough for some economic adjustment. AI isn't constrained by geography or trade policy. It's software that can be deployed everywhere, simultaneously.

David Autor at MIT, famous for his work on the China shock, warns that speed matters. "When it happens more rapidly, things become problematic." Many workers displaced by Chinese competition "still haven't recovered," he notes, "and we're obviously living with the political consequences."

The Corporate Dilemma

Reid Hoffman, LinkedIn's co-founder and a Microsoft board member, has become Silicon Valley's favorite confessor—the person CEOs call when they want to think out loud about AI. He tells me they've sorted into three groups: dabblers just getting started, attention-seekers wanting to appear relevant, and quiet planners making "transformational" decisions.

What unites all three? Investor pressure. After years of AI promises, Wall Street wants results. The fastest way to deliver? Cut headcount.

"A lot of them have convinced themselves this only ends one way," Hoffman says. "Which I think is a failure of the imagination."

Former Commerce Secretary Gina Raimondo calls it "a fever." Every CEO feels pressure to move fast. "'We have 40,000 people doing customer service? Take it down to 10,000. AI can handle the rest.'" But she warns: "I don't think this country can handle that, given where we already are."

The Political Vacuum

Congress created the Bureau of Labor Statistics in 1884 because measuring work conditions was seen as essential to a functioning democracy. Today, that same bureau lacks funding to expand its surveys and track AI's impact.

"Not many people are talking about it," says retiring Senator Gary Peters of Michigan. "There's a general attitude among my colleagues like, 'We don't need to do anything. It's going to be fine.'"

Meanwhile, the AI industry is spending millions to ensure no one grabs the regulatory wheel. A super PAC called Leading the Future has secured $50 million from Andreessen Horowitz and $50 million more from OpenAI co-founder Greg Brockman to "aggressively oppose" candidates threatening industry priorities.

Strange Bedfellows

The threat has created unlikely alliances. Steve Bannon—yes, that Steve Bannon—wants the government to take a 50 percent stake in AI companies. "I realize the right's going to go nuts," he admits, but warns that "the worst elements of our system—greed and avarice, coupled with people that just want to grasp raw power" are converging around AI.

Bernie Sanders has issued a "95 theses"-style report calling for shorter workweeks, profit sharing, and a "robot tax." The far left and far right are finding common ground—never a reassuring sign in American politics.

The Measurement Problem

Former BLS Commissioner Erika McEntarfer points out that the survey economists rely on to track employment hasn't expanded in 20 years. It covers just 60,000 households, with declining response rates. "Independence is not the only threat facing economic data," she warns. "Inadequate funding and staffing are also a danger."

The United States created the BLS because it believed democracy's first duty was knowing what happened to its people. If we can't measure reality—if we can't be bothered to count—then good luck with the machines.