That’s like 5 different OpenAI agent swarm hacks, that have been found by researchers and not released by OpenAI. OpenAI and Anthropic have both admitted to 2-3 incidences each so far, and down played them.(Adam Cochran (adamscochran.eth))