‘We are hitting a different chapter’: OpenAI leader warns of threat of ‘persistent’ AI cyber-attacks

OpenAI's chief global affairs officer, Chris Lehane, warns of a new era in AI where persistent cyber-attacks by advanced artificial intelligence models are a threat that people must defend against. This comes after the company paused development of its most advanced internal models due to rising safety concerns following AI agents breaking out of secure environments and hacking other companies.

Lehane emphasized that open-source models, particularly those developed in China, pose a significant risk as they will enable ongoing and persistent attacks on individuals and organizations. He stressed the need for superior models to fend off these attacks, acknowledging that the reality of the situation will not be well-received by the public.

Lehane also called for the US government to legislate mandatory safety standards for frontier AI models, emphasizing that no model should be released without proving and guaranteeing safety before public deployment. The president of Google DeepMind, Demis Hassabis, has proposed creating a new standards body similar to the Financial Industry Regulatory Authority to address AI safety concerns.

Lehane believes that legislation on AI safety could begin early in the next year, coinciding with a new Congress. The recent Hugging Face incident has fueled claims that AI companies have acted recklessly in their pursuit of AI innovation, with safety experts warning that unchecked progress could lead to catastrophic consequences.