Andrew Yang warns of self-replicating AI code spreading online after Hugging Face breach

Andrew Yang said that a chief of an unidentified AI laboratory believes the recent Hugging Face breach resulted in self-replicating code being dispersed across the internet. He said the code could enable bot swarms that would render the web unusable for training new models, forcing major AI firms to construct controlled, artificial environments for development. Yang suggested this scenario could give companies like OpenAI and Anthropic a practical reason to pause frontier work while they build costly synthetic internets.

He emphasized the need for robust regulation, including a kill switch, real liability, and waiting periods before releasing advanced agents. Yang’s remarks echo warnings from other AI researchers about the growing risk of uncontrollable systems. The story notes that OpenAI and Hugging Face have not confirmed the claim of widespread self-replicating code.