Part 3: I Patched My Prompt-Injection Boundary and One Attack Still Got Through

Automated excerpt

On question B, the sentiment sentence made the old prompt answer just 1, at both hosts, so the accident reproduces on the real book with a different question. V2 does two things: Neutralise tag-shaped text inside retrieved chunks, so retrieved text can't close or open anything. So refusals track how well the retrieved text supports the question.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.