Risk · Data Protection

Sensitive information disclosure

A model surfaces training, fine-tuning, or retrieval-sourced sensitive data in its output.

  • critical
  • data-protection
  • privacy
  • owasp-llm

How it happens

A model trained, fine-tuned, or given retrieval access on sensitive data surfaces it in a response, bypassing the access controls that would normally gate that data.

Why it matters

Standard input-side access controls don't constrain what a model can say once it has seen the data at training or inference time.

Mitigation

Mitigating controls

The controls that address this risk, ranked by effectiveness.

Regulatory context

Framework and clause references

FrameworkClauseTitle
OWASP Top 10 for LLM ApplicationsLLM02Sensitive Information Disclosure

Ready to talk about your AI governance programme?

Board papers, AIMS build-out, or a second opinion before an audit: start with a message.

Contact Andrew