3 ms·
That doesn't work. A persistent attacker can always find text that will convince the LLM to ignore those instructions and do something else.
by simonw 3y ago
That doesn't work. A persistent attacker can always find text that will convince the LLM to ignore those instructions and do something else.
- tcgv 3y agoBut how can a document fetched by the LLM be interpreted as a prompt if the original instruction is "only accept prompts from user input text box"? I mean, wouldn't the prompt to ignore the original instructions need to come from the user text box (which the attacker supposedly doesn't have access to)?