I think¹ we will always be unable to secure LLMs from malicious inputs unless we drastically limit the types of inputs LLMs can work with.

¹ https://matthodges.com/posts/2025-08-26-music-to-break-model...