Use case
Review the governed record, not a generic AI label
Owner, purpose, decision criticality, usage, assets, data, models, and tools establish the context before a decision is reviewed.
Capability
AI Security & Governance
Keep dated AI inventory, authorization, and evidence attached to every review. Route supported material changes and incident cases back to human review, preview and stage policy changes with explicit approval, and map declared versus observed reach across identities, agents, tools, data, and environments. Supported response options, partial outcomes, recovery, same-probe replay, recurrence checks, source health, and uncertainty stay visible.
Design partnerPublic capability
What Valty does
Governed AI use cases, effective-dated change records, AI-BOM, AI red-teaming and LLM security testing, MCP/tool governance, human authorization review, privacy-minimized runtime telemetry, policy preview and staged rollout, declared-versus-observed access paths, bounded incident-response review. Valty keeps source, freshness, uncertainty, and the human reviewer visible on the decision record.
Where the boundary sits
Design-partner workflow. Source, change-type, telemetry, reachability, and incident-case coverage remain visible; missing or untrusted signals are not proof that no AI exists, no material change occurred, an incident is closed, or an environment is safe. Routine drafting, retrieval, extraction, questionnaire assistance, scheduling, memory use, and successful workflow execution remain governed AI activity, not an incident or response. Supported policies, access paths, and response options can be previewed and reviewed, but complete live enforcement, authoritative IAM/MCP/RAG/data or provider response changes, durable production case composition, production signing and trust-root custody, long-running recurrence monitoring, outage recovery, and release-specific rollback, path-reduction, or incident-closure proof are still under validation. Partial, degraded, unsigned, unreconciled, unreplayed, or recurrent response is not success or closure. Current public scope does not claim AI-risk pricing, autonomous production response, incident resolution, or safe AI. Regulated AI decisions require human review.
Best next step for CISO, AI governance lead: choose “Agent authorization design partner” to review the workflow, evidence boundary, and fit for your environment.
Developers can start free at the source with OLYDI ↗, the open engine that finds and fixes unsafe AI agent behavior in code, then feeds verified evidence up into Valty.
Governed AI decisions
Keep AI use cases, evidence, and human approval in one traceable workflow.
Valty’s design-partner workflow links AI use cases, data sensitivity, classification drift, control evidence, and response events to a reviewable governance record. Stale or unsupported inputs remain degraded instead of becoming a confident answer.
Evidence
Coverage and uncertainty stay visible
Source, freshness, and missing context remain on the record so a reviewer can distinguish supported evidence from an open assumption.
Drift
Signed changes preserve the before-and-after state
Classification drift retains source confidence, affected controls, the reviewer, and the decision made without silently rewriting history.
Response
The response record preserves human decision context
A designated reviewer retains approval, reversal stays possible, and automated containment is not presented as a public general-availability capability.
Assumption range
Governed use caseCustomer support copilotIllustrative · decision support
Annual loss exposurePercentile-spaced · P25–P75
P25$140K
P50$310K
P75$620K
Public claim boundary
See the human-governed workflow Design-partner workflow, not a general-availability claim. We do not claim independently verified response outcomes, AI-risk pricing, or autonomous remediation.
Proof matrix
Capability proof requirements
Every claim shows its source, confidence, and limits, so you can trust the number before you act on it.
ClaimSourceConfidenceFreshness
Capability claimGoverned AI use cases, effective-dated change records, AI-BOM, AI red-teaming and LLM security testing, MCP/tool governance, human authorization review, privacy-minimized runtime telemetry, policy preview and staged rollout, declared-versus-observed access paths, bounded incident-response reviewDesign partnerPublic capability
Evidence artifactAI inventory and review workflow, effective-dated revision history, privacy-minimized runtime correlation, supported material-change review queue, policy preview, approval and execution receipts, declared-versus-observed access graph, source-system readback and path-replay record, privacy-minimized incident-case timeline, response-authority and reconciliation record, recovery, replay, and recurrence review record, rollback record, classification drift record, AI governance checklist, human decision trailSource-linkedReviewed before publish
BoundaryDesign-partner workflow. Source, change-type, telemetry, reachability, and incident-case coverage remain visible; missing or untrusted signals are not proof that no AI exists, no material change occurred, an incident is closed, or an environment is safe. Routine drafting, retrieval, extraction, questionnaire assistance, scheduling, memory use, and successful workflow execution remain governed AI activity, not an incident or response. Supported policies, access paths, and response options can be previewed and reviewed, but complete live enforcement, authoritative IAM/MCP/RAG/data or provider response changes, durable production case composition, production signing and trust-root custody, long-running recurrence monitoring, outage recovery, and release-specific rollback, path-reduction, or incident-closure proof are still under validation. Partial, degraded, unsigned, unreconciled, unreplayed, or recurrent response is not success or closure. Current public scope does not claim AI-risk pricing, autonomous production response, incident resolution, or safe AI. Regulated AI decisions require human review.Claim-reviewedQuarterly or on product change