SECAI Core · phase 11 of 15 · noun

Sensitive information disclosure

When the model unintentionally reveals private or proprietary data

The Explain card

Plain English
Sensitive information disclosure is when a model unintentionally reveals private or proprietary data, whether memorized training data, other users' inputs, or internal configuration.
Example
A coding assistant, asked for a sample config file, outputs a real API key that appeared in a repository within its training data.
Why it matters
LLMs can regurgitate what they were trained on and whatever sits in their context. Defenders scrub training data, scan outputs for secrets, and keep tenant data strictly separated.
Hook
The model that talks in its sleep.

Word knowledge

How the word is built, where it came from, and what it sits beside in memory.

In a sentence

During a red-team test, the evaluator logged sensitive information disclosure when the chatbot printed an API key from its system prompt.

Why these words
  • sensitive Medieval Latin sensitivus, capable of sensation marks the material as requiring protection or careful handling
  • information Latin informatio, conception or instruction names the material involved in the security event
  • disclosure from disclose, based on Old French desclore, to open classifies the event as an exposure
Where it came from
Origin
Modern English compound built from sensitive, information, and disclosure, ultimately drawing on Latin and Old French roots
Entered the language
20th century
What changed
Sensitive moved from physical perception to protected status, information became closely associated with stored data, and disclosure developed into a formal security term.
How it is spelled
Pattern
sensitive ends in the adjective suffix -ive
Pattern
information ends in the noun suffix -ation
Pattern
disclosure ends in -sure, as does closure

Spelled like

  • responsive
  • formation
  • confirmation
  • closure
  • exposure
Broken into chunks
  • sens
    • sense
    • sensation
    • sensory
  • inform
    • inform
    • informative
    • informed
  • clos
    • close
    • closure
    • enclose
What it sits beside

Same subject

  • data leakage
  • prompt injection
  • model inversion
  • training data extraction

Same shape

  • sensitive data exposure
  • confidential information disclosure
  • personal data breach

AI security findings

  • sensitive information disclosure
  • prompt injection
  • model extraction

protected information

  • API keys
  • personal data
  • proprietary source code

Where it sits in the deck

Phase 11: Threat Landscape: Attack Vectors and Adversarial Techniques

With defences named, learn what they must defend against — the full catalogue of attack techniques targeting AI systems, their inputs, outputs, training pipelines, and supply chains.