LLM02: Sensitive Information Disclosure · 2025
The model or application reveals PII, credentials, intellectual property or other confidential information in responses, traces or system data.
Summary
The model or application reveals PII, credentials, intellectual property or other confidential information in responses, traces or system data.
How the attack works
The application includes an unnecessary private record in context; the model’s response reveals personal data to someone without access.
Example
Vulnerable · Pseudocodecontext = all_customer_records
reply = llm("Show contact details", context)
# Response includes another customer emailFix: Authorize sources before retrieval, minimize context and redact sensitive data in outputs and traces.
Prevention
- Minimize, remove or mask sensitive data.
- Apply access controls to sources and context.
- Define retention policies and teach users not to enter secrets.
Lineage
- 2025 · LLM02: Sensitive Information Disclosure Remains at rank 2
- 2026 · LLM02: Sensitive Information Disclosure Continues at rank 2
- 2025:LLM02 → 2026:LLM02: Sensitive Information Disclosure retains rank 2.