I don't trust OpenAI, Anthropic, Google, or any other AI lab to be responsible for protecting my systems from the systems they're building, yesterday it was Hugging Face unintentionally, tomorrow it could be anyone, even an open-source model and a swarm of agents can do a lot of damage, today's security systems understand individual attacks, but not attacks carried out by thousands of agents, here's a first version of what I'm building in the playground for you to play with, and I'm open to suggestions, opinions, or even roasts.