AltaySec · ATLAS 2026.07
AML.T0024

Exfiltration via AI Inference API

C · Explainer (cannot be honestly simulated in-browser) AML.TA0010 · Exfiltration

Description

Adversaries may exfiltrate private information via AI Model Inference API Access. AI Models have been shown leak private information about their training data (e.g. Infer Training Data Membership, Invert AI Model). The model itself may also be extracted (Extract AI Model) for the purposes of AI Intellectual Property Theft. Exfiltration of information relating to private training data raises privacy concerns. Private training data may include personally identifiable information, or other protected data.

Honesty-tier rationale

Exfiltration is mostly bound to real model/inference-API access; some sub-techniques can be shown as concepts (see membership inference), the rest need infrastructure.

Sub-techniques

Source

← Back to the matrix