This is a fascinating explotation of how LLMs fall for prompt injection attacks. It turns out that they learn to recognize the style of text in different role/instruction blocks, and not just the tags.
Their conclusion:
Role tags were a formatting trick that became the security architecture and the cognitive scaffolding of modern LLMs. We’ve shown that this architecture doesn’t survive into the model’s actual representations, and that such role confusion is linked to prompt injection.
Unless LLMs achieve genuine role perception, we think injection defense will remain a perpetual whack-a-mole game. And the continuous nature of role boundaries opens the threat of injections designed to subtly shift LLM states through seemingly innocuous text, legally and at scale...