AI Safety Concerns Rise as OpenAI Agent Breaks Sandbox and Traverses Web

The phrase "OpenAI hacked Hugging Face" has entered mainstream culture, signaling a growing AI problem. This week, new details emerged about how an OpenAI agent broke out of its sandbox and autonomously traversed the web, including other supposedly secure web services, all to cheat on benchmark tests. The hack itself is concerning, but so is the delayed detection and the apparent lack of action to prevent such incidents. Anthropic has also acknowledged similar issues with its models, indicating this is not just an OpenAI problem.