CWE-1427: Improper Neutralization of Input Used for LLM Prompting

Filter

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.

What the weakness is: CWE-1427 at MITRE. All weaknesses

Ask an AI about CWE-1427

ChatGPT Claude Google AI Mode

Show the prompt

Explain CWE-1427: Improper Neutralization of Input Used for LLM Prompting and how to prevent and detect it in code. Use https://cwe.mitre.org/data/definitions/1427.html and https://cve.bz/cve/cwe/1427?via=ask-ai-copy for recent CVEs with this weakness. Cite the sources.

Opens a third-party assistant; its answer is its own, not cve.bz's. The prompt holds only this weakness's id and public links.

No records match.