19:38
2026-07-30
pantsyr.dev
large-language-models
Why prompt injection is still possible in LLM applications
Prompt injection remains possible in large language model applications because models cannot distinguish between trusted instructions and untrusted user input in the token stream, according to a techn…