Gebru said claims that AI poses an existential threat are not just a distraction but a harmful regulatory-capture strategy that steers governments away from policing current abuses.
She argued the same investors and founders behind leading AI firms also fund safety institutes and auditors, letting companies market “superintelligence” to investors and states while shaping the rules meant to govern them.
At the UN and in earlier policy fights, she said, executives such as Sam Altman and Dario Amodei warned of global AI risk even as companies resisted binding oversight—citing threats to leave the EU or lobbying to weaken rules.
Gebru said existing law already covers much of the problem, pointing to Lina Khan’s framework: deceptive marketing, data transparency and documentation, labor exploitation, and limits on data theft.
She also defended her 2021 “stochastic parrots” research as still applicable to today’s chatbots, arguing large language models still generate likely text patterns rather than genuine reasoning, which fuels overtrust and real-world errors.