How to meet it
OWASP AISVS v1.01
Adversarial Robustness · Model-Extraction Defense
Don't expose raw model outputs
Verify that raw model outputs are not directly exposed beyond the application backend, and that externally visible responses are calibrated to the extraction risk level.
OWASP AISVS v1.01, requirement 11.3.2
In plain English
Keep raw model outputs such as full probability scores inside your backend, and return only what the use case needs.
What good looks like
- External responses contain only the information needed.
- Probabilities and logits are not exposed unless required.
- The level of detail is set according to extraction risk.
How teams usually do it
- Returning labels or rounded scores instead of full distributions.
Evidence an auditor would accept
- API response schemas.
Common mistakes
- Returning full log-probabilities because the default API did.
Not sure which requirements apply to your system? Take the assessment.
Requirement text from the OWASP AI Security Verification Standard, licensed under CC BY-SA 4.0. This guidance is shared under the same licence.