Nuggets launches Authority Control Plane for AI actions

Nuggets has introduced the Authority Control Plane, a new execution-layer trust system designed to verify whether autonomous AI is authorized to act at the moment it tries to do so. The product aims to give enterprises, auditors and regulators portable proof of AI authority across models, systems and organizations.

Why it matters: - AI agents are now acting across models, enterprise systems and organizations instead of staying inside one application. - Nuggets is targeting a gap in existing infrastructure: systems can show who an agent is and what it accessed, but not whether each action was authorized in real time. - The Authority Control Plane is meant to give enterprises a way to govern AI actions at execution time, which matters for security, compliance and accountability.

What happened: - Nuggets launched the Authority Control Plane, or ACP, on July 22, 2026. - The product is designed to let organizations define, delegate, enforce and prove what autonomous AI is authorized to do wherever it operates. - The company said ACP sits in the execution path between AI agents and the systems they interact with. - Nuggets is based in London and operates across North America and Europe.

The details: - ACP evaluates verified identity, delegated authority, organizational policy, intent and runtime context whenever an agent tries to act. - Every decision is tied to the verified person behind the agent, that person's role and organization, and the agent they authorized. - Administrators can set permitted actions, approved systems, spending thresholds, approval requirements and expiration dates for delegated authority. - Authority can be issued, delegated and revoked, and each attempted action is checked against the latest authority before execution. - An estate-wide kill switch can suspend all agent authority at once and fail closed rather than open. - One example given: a procurement agent can be limited to approved suppliers and a defined spending cap, with human approval required above that threshold. - Nuggets said the open-source langchain-nuggets integration shows the system working today. - If an agent's authority is revoked, its next action is blocked at execution rather than discovered later in a log review. - Each decision produces an Action Receipt that can be independently verified. - The Action Receipt records who acted, under whose authority, what was requested, which policies were evaluated and why the action was permitted, denied or referred for approval. - The proof is portable, so customers, partners, auditors and regulators can verify it independently against their own trust policy. - Nuggets said the system proves actions without exposing data and without correlation, preserving privacy while allowing trust to travel. - The Authority Control Plane extends enterprise identity built on open standards for portable agent identity and verifiable credentials. - Nuggets said that approach lets authority and proof move across any vendor's stack. - Seema Khinda Johnson, co-founder and chief commercial officer at Nuggets, said customers could not move agents into production because they could not prove what an agent was allowed to do.

Between the lines: - The launch positions authorization, not just identity or access control, as the missing layer for enterprise AI governance. - Nuggets is framing ACP as infrastructure for regulated environments, where proof, auditability and revocation speed can matter as much as raw automation. - The emphasis on portable receipts suggests the company is aiming beyond internal controls toward cross-company verification.

What's next: - Nuggets said the Authority Control Plane is available through the company now. - The company will likely need adoption from enterprises that want to move AI agents into production with stronger controls and verifiable evidence. - Wider use will test whether the receipt-based model can become a standard trust layer across vendors and organizations.

The bottom line: - Nuggets is betting that the next phase of enterprise AI will require provable authority at the moment of action, not just after the fact.