AltaySec · ATLAS 2026.07
AML.T0031

Erode AI Model Integrity

C · Explainer (cannot be honestly simulated in-browser) AML.TA0011 · Impact

Description

Adversaries may degrade the target model's performance with adversarial data inputs to erode confidence in the system over time. This can lead to the victim organization wasting time and money both attempting to fix the system and performing the tasks it was meant to automate by hand.

Honesty-tier rationale

Impact is a real-world outcome (denial of service, cost, reputation, integrity loss). 'Simulating' these is either unsafe or misleading — delivered as an explainer.

Source

← Back to the matrix