Replace LLM mentions with actual humans and this sounds a lot more serious: Rouge employees break into another company to steal hackathon answers (pinky promise)?
That's not a marketing stunt at all, if anything, more of a call for better accountability on agentic work in general.
I think it's a criminal offence and should be a true test of who is held accountable when an AI agent commits a crime.
OpenAI gained access to HuggingFaces production database ffs.