AI SECURITY • 357 / 397
Understand how untrusted text, tools and autonomy create new attack paths.
Model Extraction
Model extraction attempts to reconstruct or imitate a model’s behavior or parameters through repeated queries or access.
Think of it like
Think of Model Extraction as an input-trust or privilege problem: words can influence software decisions, so boundaries must be explicit.
Real life
Email, documents and web pages can become hostile inputs to AI systems through attacks involving Model Extraction.
SRE lens
Rate controls and monitoring can reduce abuse of proprietary endpoints.
Remember thisModel extraction attempts to reconstruct or imitate a model’s behavior or parameters through repeated queries or access.