AI Job Shock Reshapes Who Works, Who’s Paid, and Who’s Invisible

AI isn’t just automating tasks; it’s becoming the official reason people lose their jobs, even when the numbers don’t add up. U.S. employers attributed more than 54,000 layoffs to artificial intelligence in 2025, according to outplacement firm Challenger, Gray & Christmas, but economists argue many firms are “AI‑washing” broader cost‑cutting and post‑pandemic over‑hiring problems to placate investors. Tom’s Hardware and other analysts say AI is often a convenient scapegoat, even as a New York Fed survey found only 1% of service firms reported layoffs directly due to AI in late 2025 and many actually hired or retrained workers instead. Vygr News reports similar findings.

Yet early‑career workers are on the front line of the risk. Anthropic CEO Dario Amodei has warned that AI could wipe out up to 50% of entry‑level white‑collar jobs and push unemployment toward 20% within five years, a projection Nvidia’s Jensen Huang publicly disputed as overly pessimistic. Tom’s Hardware covered the clash, while a U.S. senator recently warned that graduate unemployment could reach 25% as companies automate routine analysis and support work, risking “unprecedented” social disruption if the career ladder’s first rungs disappear. Fortune linked those fears to already‑rising joblessness among recent grads.

The shock is compounded by who actually powers AI systems. Investigations show OpenAI and others outsourced toxic‑content labeling to Kenyan workers earning under $2 an hour via contractor Sama, despite contracts billing at several times that rate, leaving annotators traumatized enough to petition their government over “exploitative” conditions and form Africa’s first content‑moderator union. Time and The Guardian documented how this hidden workforce underpins “ethical” AI. New research on Africa’s data‑annotation gig economy finds similar pay levels and weak legal protections even as the work becomes central to global AI supply chains. CIPIT.

At the other end of the spectrum, improv actors and performers are being recruited to teach AI systems emotional nuance. A recent “improv actor – AI trainer” fellowship advertised by Handshake AI asks performers to craft character, subtext and feeling so models can sound more human in future products, turning affective labor into another narrowly scoped, project‑based gig. Handshake describes the role as a “paid, collaborative” experiment, but offers few clues about residuals or long‑term credit once their performances are baked into commercial systems.

Taken together, AI is compressing the middle of the labor market—threatening entry‑level white‑collar roles while leaning on precarious, often low‑paid workers to make models safer and more persuasive. The technology’s real disruption may be less about total job extinction than about who holds bargaining power, who is visible on the payroll and who is quietly written out of the social contract.