Constitution Runtime (Design)
DevelopmentProposed five-dimension constitutional review model. Current public evidence is demonstrative; enforcement coverage and drift resistance require a separate code-level audit.
View Demo ArtifactAI governance & trust · UFA in development
Governance and evidence
for AI agents
AIF Trust develops authorization, approval, and evidence controls so organizations can govern what AI agents do. UFA (Unified Frontier Agent) is the agent control plane we are building to apply those controls. It is in development and not production-ready.
UFA: Partially Certified
Acceptance tests: evidence-stamped
Demo evidence: not verified
Security acceptance: not ready
AIF Trust and UFA
The organization
AIF Trust
Focused on AI governance and trust: defining how agent actions are authorized, approved by people where needed, and recorded as reviewable evidence.
The software
UFA — Unified Frontier Agent
The agent control plane AIF Trust is building: task engine, approval gates, and a signed evidence ledger. It is a development project, not a released product.
UFA status
- Certification
- PARTIALLY_CERTIFIED
- Stage
- In development
- Production readiness
- Not production-ready
Security acceptance is NOT READY. No customer deployments are claimed, and no independent audit has been completed. Details and known limitations are on the UFA page.
Request Early AccessTechnical Capabilities (Demonstration)
Each capability below has a demonstration evidence package. Independent verification pending — no production claims without audit.
Durable Execution Design
DevelopmentProposed durable-execution path covering retries, compensation, scheduling, signals, and replay. Actual coverage and Temporal integration must be verified per deployment.
View Demo ArtifactEvidence Package Format (Demo)
DevelopmentEvidence-package format under validation. The current public JSON examples contain demonstration placeholders and are not verified execution evidence.
View Demo ArtifactIdentity & Attribution Design
DevelopmentIdentity attribution and end-to-end provenance are design goals. Enforcement must be demonstrated and independently tested before production claims.
View Demo ArtifactPolicy Decision Point Design
DevelopmentPolicy decision-point design based on explicit allow/deny rules and OPA/Rego. Platform-wide production enforcement is not established by the current demonstration artifacts.
View Demo ArtifactContinuous Audit (Prototype)
DevelopmentPrototype direction: discovery, drift detection, proposed remediation, and evidence reporting. Automation scope and observed outcomes still require validation.
View Demo ArtifactUnified Memory Fabric (Preview)
PreviewPlanned memory layers include working, skill, episodic, and semantic memory. Autonomous consolidation and tiering require separate implementation evidence.
View Demo ArtifactMission Scheduler (Development)
DevelopmentMission-orchestration concept under development. UFA has historical Wave 19 evidence of 12/12 acceptance missions on its stamped code SHA; this does not establish live AIF missions or an integrated production deployment.
View Demo ArtifactArchitecture: Trust by Design
AIF Trust is designed around linking intent, authorization decisions, execution records, and provenance. Public evidence currently remains demonstration-only; end-to-end enforcement claims require a fresh audit and independent verification.
Intent Layer
- Intent Compiler (natural language → executable spec)
- Mission Proposal & Constitutional Review
- Human/Policy Approval Gates
Core
Execution Layer
- Temporal Workflows (durable, replayable)
- Constitutional Interceptors (auto-context)
- Activity Heartbeats & Compensation
Trust Layer
- Evidence Packages (immutable, sealed)
- Receipt Verification & Replay
- 13-Dimension Health Observatory
Unified Frontier Agent (UFA)
AIF Trust's Flagship
Agent Runtime
Unified Frontier Agent (UFA) is AIF Trust's flagship agent-runtime project. Its latest accepted evidence baseline is Wave 19; security acceptance remains NOT READY. The public evidence files on this website are demonstrations, not proof of an integrated production deployment.
12/12
Historical Acceptance Missions
312
Historical Tests Passed
Wave 19
Latest Accepted Evidence Baseline
NOT READY
Security Acceptance
INITConstitution Runtime v1.0 loaded (demo)
READY7 constitutional services configured (demo)
MISSIONtemporal-optimization scheduled 05:00 UTC (demo)
MISSIONopportunity-ranking running hourly (demo)
EVIDENCEDemonstration package sealed
VERIFYIndependent verification pending
LEARNOutcome → routing → policy → automation (demo)
DEMODemonstration mode — not live autonomous operation
Evidence Status: Demonstration
Evidence packages exist but contain demonstration data. Independent audit required for production verification.
8
Evidence Packages (Demonstration)
0%
Independently Verified
—
Awaiting Independent Audit
Ready to Prove What Your Agents Did?
Apply to discuss a bounded design-partner evaluation. Current evidence artifacts remain demonstrations pending independent verification; no production guarantee is implied.