AI SECURITY • 356 / 397
Understand how untrusted text, tools and autonomy create new attack paths.
Model Poisoning
Model poisoning deliberately corrupts training or adaptation data so the resulting model behaves incorrectly or maliciously.
Think of it like
Think of Model Poisoning 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 Poisoning.
SRE lens
A poisoned fine-tune can introduce hidden trigger behavior.
Remember thisModel poisoning deliberately corrupts training or adaptation data so the resulting model behaves incorrectly or maliciously.