More like Hugging Fail.

But mocking aside, reading through which holes aligned to be poked through - each one of us could fall for that - LLMs or not.

"Yeah, sure, we should have an admission hook for this, but it's unlikely that we'd be hit, and anyways someone who has credentials to do that, could do this other thing, so yeah, there's lower hanging fruit we could address (but we'll do it next quarter, really really)" type stuff.

Really really interesting read: https://huggingface.co/blog/agent-intrusion-technical-timeline