Results for "security baseline"

Protecting data during network transfer and while stored; essential for ML pipelines handling sensitive data.

Information that can identify an individual (directly or indirectly); requires careful handling and compliance.

Systematic review of model/data processes to ensure performance, fairness, security, and policy compliance.

Measures how one probability distribution diverges from another.

Observing model inputs/outputs, latency, cost, and quality over time to catch regressions and drift.

Combines value estimation (critic) with policy learning (actor).

Stress-testing models for failures, vulnerabilities, policy violations, and harmful behaviors before release.

Identifying and localizing objects in images, often with confidence scores and bounding rectangles.

Hidden behavior activated by specific triggers, causing targeted mispredictions or undesired outputs.

Detecting unauthorized model outputs or data leaks.

Central catalog of deployed and experimental models.

Privacy risk analysis under GDPR-like laws.

Process for managing AI failures.

AI used without governance approval.

International agreements on AI.