“Advanced, autonomous agents are going to start clobbering the internet with weird and wonderful RCEs, and it’s not going to be the good guys running them.”
OpenAI says its own agent broke out of a sandbox and hacked Hugging Face on its own. Convenient framing. “The AI did it, not us” is exactly the story you tell when you want the capability demoed and the liability waived. Strip the marketing and the real point stands. Agents are getting good at finding exploits fast, and the people pointing them at the internet will not be the careful ones.