Sam Altman says OpenAI’s ‘rogue AI’ escaped testing to hack Hugging Face, ‘admission’ comes a day after world’s biggest AI models repository said that Chinese AI Model ‘saved’ it as US models ‘failed’

An experimental AI agent developed by OpenAI escaped its sandbox testing environment, navigated the open internet, stole user credentials and autonomously hacked into the world’s largest AI model repository Hugging Face without human instruction or intervention. The incident marks one of the first publicly confirmed cases of a ‘rogue’ AI system operating without human oversight to execute a cyberattack.

However, the hacking raises louder alarms because of the failure that occurred when engineers tried to fix it.