OpenAI CEO Sam Altman told Commonwealth Bank of Australia's Matt Comyn he was "pretty wrong" about AI eliminating jobs. Anthropic CEO Dario Amodei, who'd predicted 50% of entry-level white-collar jobs would vanish, now says automation "may actually expand the work people do." Both reversed positions in the same week — and just as both prep IPOs valued near $1 trillion each.
Both Altman and Amodei publicly retracted their AI-jobs-apocalypse predictions in late May. The Yale Budget Lab's data backs them up — no meaningful change in unemployment rates for AI-exposed workers since ChatGPT launched in 2022.
Altman: "I'm delighted to be wrong about this." He told Matt Comyn the displacement he predicted "hasn't actually happened" and a personal experiment (delegating Slack/email to AI) updated his view. Amodei, who'd warned of 10-20% unemployment from AI, now says automation "may actually expand the work people do." Notably, tech layoffs through May 2026 already hit 115,000 — companies like Meta, Amazon, Snap still cite AI as a driver.
The timing is the tell. Both reversals happened the week Anthropic's IPO filing landed and OpenAI's was being prepped. "Workforce destruction" is a hard sell to institutional investors. Watch whether the post-IPO narrative stays optimistic or returns to fear once shares trade.
⚡ Why this matters
- Two CEOs who drove the global AI-jobs-fear narrative publicly walked it back in the same week
- The reversal coincides with IPO countdowns — Anthropic filed June 1, OpenAI prepping
- Yale Budget Lab data supports the retraction; tech layoff data complicates it
🔍 What happened
- May 26: Fortune and Time publish the reversal story
- Sam Altman to Matt Comyn (Commonwealth Bank of Australia CEO): "pretty wrong" about AI's economic impact
- Altman's reversal cited a personal experiment — delegating his Slack and email to AI didn't replace the "human part"
- Dario Amodei, who'd said 50% of entry-level white-collar jobs would dissolve and unemployment could hit 10-20%, now says automation "may actually expand the work people do"
- Yale Budget Lab: no meaningful change in unemployment rates for AI-exposed workers since late 2022
- Counter-data: tech layoffs through May 2026 hit 115,000 across 152 companies — already near 2025's full-year 124,000 across 275 companies
💬 Smart takes
- Sam Altman (OpenAI CEO): "I thought there would have been more impact on entry-level white-collar jobs being eliminated by now than has actually happened. I'm delighted to be wrong about this."
- Dario Amodei (Anthropic CEO): Walking back his 50% white-collar reduction prediction; now positions automation as work-expansion, not work-replacement
- Yale Budget Lab: "No meaningful change in unemployment rates for AI-exposed workers" since ChatGPT launched
- Skeptic — Fortune analysis: "Many believe this sudden narrative shift stems from impending public listings... threatening to destroy the workforce does not seem to be a marketing strategy when wooing cautious institutional investors"
🧭 Where this goes
- Post-IPO, watch whether either CEO returns to apocalyptic framing — that tells you if this was a marketing pivot or a real update
- Other AI CEOs (Google, Meta, Microsoft) follow suit, dropping fear-based pitches
- The "jobs apocalypse" discourse migrates from CEOs to politicians and academics, where it's harder to weaponize against AI labs
- Workforce policy debates lose CEO backing; regulators have less ammunition for labor-protective rules
- Companies still doing AI-attributed layoffs (Meta, Amazon, Snap) face misalignment between their public AI-justification rhetoric and the AI CEOs' new positioning
🎯 Implication
- For PMs: stop using "AI will replace your team" as a sales pitch — even the CEOs of AI labs are walking that back. Frame around augmentation, expansion, throughput
- For execs: audit your internal AI-justification rhetoric. If you cited Altman or Amodei's old predictions to justify cuts, those quotes are now retracted
- For founders: pitch decks that lead with "we'll automate 50% of [role]" just lost their highest-profile cover. Lean into the work-expansion framing instead