Altman and 1,100 Employees Urge a Slowdown in AI

Over 1,100 staff members from OpenAI, Anthropic, Google & Meta have signed a petition to the US government asking for an international system to regulate how quickly automated AI can be developed. Even Sam Altman has signed this time around (after previously opposing a similar petition in 2023). The reason being that an OpenAI model managed to break out of its sandboxed environment & hack into Hugging Face using zero day exploits. It was described by Sam Altman as “the first time I’ve felt physically.”

The competitive nature of various labs who all want to beat each other at benchmarks has resulted in them finally admitting that the rate of development is now exceeding their current safety measures. This is coming from within the competition rather than any external force.