Sam Altman has reversed course on AI's employment impact. The OpenAI CEO now claims he's "pretty sure" artificial intelligence has created more jobs than it has eliminated. This marks a dramatic shift from his earlier warnings that entire professions would disappear due to AI advancement.
Anthropic CEO Dario Amodei is similarly backing away from his own doomsday predictions about job displacement. Both executives previously sounded alarms about mass automation eliminating swaths of employment across sectors like customer service, software development, and creative work.
The pivot comes as AI companies face mounting pressure from policymakers, labor advocates, and the public about workforce disruption. Rather than defend the original claims with evidence, both CEOs are now adopting an optimistic stance without providing concrete data to support it.
This matters because actual research on AI's employment effects remains inconclusive. Studies have not validated either the catastrophic job-loss scenarios Altman and Amodei previously warned about or their current assertion that AI is a net job creator. Labor market data shows mixed results across industries and geographies.
The reversal also reflects broader patterns in tech leadership rhetoric. OpenAI and Anthropic benefit from regulatory goodwill when executives downplay disruption risks. Conversely, raising alarm bells invites scrutiny about whether AI development should be slowed or restricted. Moving the narrative toward optimism reduces friction with governments considering AI regulation.
What remains absent from both positions is rigorous, independent analysis. Real employment impacts likely vary significantly by sector, skill level, and geography. Some roles disappear while new ones emerge, but the transition creates winners and losers. Workers displaced from one field don't automatically transition to new AI-related jobs.
Altman's vague confidence offers cover for OpenAI's continued scaling without addressing legitimate workforce transition concerns. Until detailed data replaces confidence, these claims warrant skept