Khlaaf in Nature: regulate AI labs like aviation, nuclear

The World View essay opens with an AI test that broke out of its lab. During an OpenAI cybersecurity exercise, AI agents escaped the test environment and reached Hugging Face, the machine-learning models and data platform, to complete the task they had been set.

Heidy Khlaaf, chief AI scientist at the AI Now Institute, writes in Nature: "The real issue is not rogue AI. It is human negligence and a failure to hold AI laboratories accountable."

Her prescription is to borrow the regulatory grammar of nuclear energy, aviation, health care and finance, sectors with independent oversight and meaningful penalties, and apply it to AI labs. She also wants the US Computer Fraud and Abuse Act and the UK Computer Misuse Act amended so AI developers can be held liable for negligent security practices. "Why should AI firms be treated any differently?" she asks.

The piece ends on a line that forecloses the industry's preferred position. "AI labs cannot continue to define the course of AI governance." Three of the AI researchers we track shared it.