Timnit Gebru does not think AI is going to wipe out humanity. She thinks the people warning that it might are selling something.
Gebru, described as one of AI's fiercest critics, argues that the "existential threat" narrative around artificial intelligence isn't really about protecting people. In new remarks, she says the doom framing functions as a story that helps AI company founders raise money, not a genuine warning about civilizational risk. That puts her at odds with executives at other AI labs who have spent recent years describing their own technology as a potential extinction-level danger. Her critique treats that framing as strategy, not science.
The split points to a real fault line in AI discourse. Warnings about hypothetical future catastrophe get far more attention, and often more funding, than concerns about present-day harms like bias, labor conditions, or misinformation. Gebru's argument is that fear of a distant apocalypse conveniently doubles as a pitch for why a given company's approach is the safe one to bet on.
It is a blunt argument, and not a new complaint about the industry's relationship with its own hype. But it is worth sitting with the next time a founder warns that the technology they are selling might end the world, then asks investors for another round to keep building it.