OPAQUE Launches Verifiable AI Governance Standard

The platform introduces two primary open-source milestones: Agent Manifest and Confidential MCP. Agent Manifest serves as a new standard for runtime integrity, ensuring that organizations can cryptographically verify an agent's identity and its authorized actions. Complementing this, Confidential MCP provides the first Model Context Protocol implementation secured through confidential computing, enabling hardware-enforced policy execution for every tool call an agent performs.

OPAQUE 3.0 addresses the growing regulatory and security demand for automated, hardware-signed evidence of AI behavior. By leveraging CPU-based confidential computing from AMD, Intel, and NVIDIA, the system allows enterprises to audit AI workflows without needing to trust the underlying service provider. This approach builds upon the Agent Governance Toolkit, an open-source framework created by Chief Platform Officer Imran Siddique, who now leads the development of these new verifiable standards.

To ensure long-term security, OPAQUE is partnering with the Technology Innovation Institute to incorporate post-quantum cryptography into the platform. This preparation aims to keep audit records and agent identities verifiable against future quantum-era threats. The company plans for general availability in July 2026, with the open-source components currently accessible via the agentrust-io repository.