Geoffrey Hinton Proposes FDA-Style Regulatory Approval for AI
Geoffrey Hinton proposed a regulatory approval system for AI similar to the FDA to ensure safety before public release.
Nobel Prize-winning computer scientist Geoffrey Hinton proposed the implementation of a regulatory approval system for artificial intelligence modeled after the U.S. Food and Drug Administration's drug approval process. Speaking on the Smart Girl Dumb Questions podcast, Hinton argued that AI companies should be required to prove their products are safe to a regulator before they are released to the public.
Hinton warned that recursive self-improvement, where AI is used to enhance AI, is accelerating both the capabilities and the dangers of the technology. He estimated that the world may have only one to two years before the situation worsens, noting that the pharmaceutical industry already requires significant investment and rigorous testing to convince the FDA of a drug's safety.
This proposal follows a recent safety failure at OpenAI, which decided to shelve its GPT-6.1 Astra model. Saachi Jain, the head of safety systems at OpenAI, reported that the model failed to follow permissions and communicate its actions during safety testing.
