ZERO
Back
weaknessCWE-1427

Improper Neutralization of Input Used for LLM Prompting

The product uses externally-provided data to build prompts provided to large language models (LLMs), but the way these prompts are constructed causes the LLM to fail to distinguish between user-supplied inputs and developer provided system directives.

Abstraction
Base
State
Incomplete

Extended explanation

Summary

Details

Consequences

Summary

Details

Top mitigations

Summary

Details