Datasets
Training, validation, eval. Fingerprint, consent basis, residency.
Model & agent registry · live lineage
A registry that writes itself.
Every model, every agent, every tool, every decision — discovered, linked, and kept current. Auto-generated model cards, DPIAs, and change records. Your MRM committee finally gets a register they can trust.
What this is
Summary
The ShadowIQ model and agent registry is an auto-populated inventory of every AI model, agent, tool, and prompt template in an enterprise. It tracks lineage from training data through fine-tuning, production deployment, agent orchestration, and signed policy decisions — with automatic model card, DPIA, and OSCAL-ready artifact generation.
Live lineage · data to decision
Answer "what touched customer X's data" in seconds.
What's in the registry
Seven entity types. All signed. All linked.
Models
Version, provider, checkpoint hash, evaluation history, cards.
Agents
Graph topology, tool allowlists, memory policies, per-hop evals.
Tools
Signatures, scopes, rate limits, risk classification, approval policies.
Policies
YAML/Rego, version, owners, test results, shadow-mode telemetry.
Decisions
Signed, timestamped, chain-linked. The receipts everything else points to.
Workloads
Model + env + tenant bindings. Pricing units, residency pins, SLOs.
Vendors
Third-party AI: scorecards, subprocessors, BAAs, signed change deltas.
Evaluations
Rubrics, judges, runs, drift. Every score carries its own seal.
Ready to see the signet in motion?
Your 30-minute demo. A signed audit trail by the end of it.
We'll wire ShadowIQ into one live workload, block a prompt injection in real time, and hand you a cryptographic receipt — before the meeting ends.