Capability · 01
Supply-chain integrity
Pre-training corpus integrity · fine-tuning data lineage · weights provenance · inference-pipeline verification.
What this module does
Vulnerability scanning across the AI supply chain, adversarial robustness testing, model access control, and third-party AI risk assessment. The defensive perimeter around your models, agents, and MCP servers, not just the runtime.
Capabilities
Four operational capabilities.
Capability · 02
Adversarial robustness
Robustness testing · red-team scenarios · membership-inference defense · model-extraction defense.
Capability · 03
Access control
Per-agent model access control · scoped invocation · audit trail.
Capability · 04
Third-party AI risk
Vendor-risk assessment for third-party AI APIs · contractual no-training verification · regional data flow.
Technical detail
What evaluators want to know.
Robustness scoring
Per-model robustness scores across prompt-injection, data-poisoning, membership-inference, model-extraction.
Supply-chain stages
Pre-training corpus → fine-tuning data → model weights → inference pipeline · status per stage.
Access policy
Cryptographic agent identity · scoped invocation policy · audit-grade trail.
Vendor-risk reports
Per-vendor compliance posture · regional data flow · contractual terms cited.