DyFram Govlanes resolves an AI system's risk profile into a jurisdiction-committed G-Level — with obligations, red-bands, and prohibition enforcement — verifiable end to end.
A structured, seven-dimension questionnaire — capability, use, impact, control, context, actor, and ownership — captures how an AI system actually behaves.
Critical Prohibition Patterns are checked first. A match halts resolution immediately, with a clear, jurisdiction-aware explanation — before any score is even computed.
A deterministic, three-stage pipeline resolves the seven dimensions into a single risk score, using jurisdiction-committed scoring configuration — never a black box.
The score resolves against the jurisdiction’s own committed threshold bands, with red-band rules able to raise the level when specific patterns are present.
The resolved G-Level pulls in the jurisdiction’s obligations and Jurisdictional Adaptation Layer — the concrete, local requirements a system now has to meet.
A Governance Core Package is assembled with the submitter — never stored in full by the platform — and a public package code anyone can independently verify.
Committing a JAL version, publishing thresholds, activating a red-band or CPP rule, onboarding a government workspace — every major action sits pending until a second teammate approves it. A single reject halts it.
A government workspace can only ever act on its own jurisdiction’s data. Scope is derived from server-side membership on every request — never accepted from a client payload.
Every resolved assessment gets a package code. Anyone — a regulator, a partner, a curious citizen — can check its validity, provenance, and G-Level without an account.
Manage jurisdictions, questionnaire, engine configuration, and government onboarding across the platform.
Manage your jurisdiction's JAL, thresholds, red-bands, CPP rules, and obligations.
New here? Create an account — you'll need a workspace invitation to reach a dashboard.