DigiCert Launches AI Trust Framework to Secure Autonomous Enterprise Agents

DigiCert has introduced "AI Trust," a governance framework designed to secure the deployment of enterprise AI agents and models. The initiative aims to solve the growing security gap created by autonomous AI workloads that operate outside traditional human-centric security controls.

At the core of the framework is the "AI Agent Passport," a cryptographically protected record that links an agent's identity directly to its authorization. These passports contain critical metadata, including permitted operations, approved systems, data sensitivity classifications, expiration dates, and the identity of a named human owner. To manage these identities, the framework utilizes SPIFFE (Service Profile Identity) and SPIRE (SPIFFE Registration Endpoint), treating AI agents as workload identities rather than extensions of human Identity and Access Management (IAM).

The Rise of Shadow AI

The push for agentic AI has outpaced corporate security infrastructure, leading to a surge in "shadow AI." Brian Trzupek, Senior VP of Product at DigiCert, noted that when the promise of the technology is high, organizations are often willing to "throw security out the window" to achieve rapid deployment. This urgency often results in the use of insecure, long-lived static API keys, which provide a broad attack surface for malicious actors.

This risk is underscored by an IBM 2026 Cost of a Data Breach report, which found that 68% of organizations lacked the necessary governance to manage or detect shadow AI within their environments. DigiCert is leveraging its background in Public Key Infrastructure (PKI) and DNS to move enterprises toward a verifiable trust architecture for these non-deterministic workloads.

Hardening Model Integrity

Beyond identity, the framework addresses the integrity of the AI models themselves. DigiCert advocates for a layered defense involving OCI-compliant packaging and Sigstore code signing. To prevent tampering during execution, the framework recommends the use of trusted execution environments (TEEs), specifically citing Intel TDX or AMD SEV-SNP.

Furthermore, DigiCert proposes using DNS as a policy enforcement point. By implementing a DNS-based policy record—similar to the DMARC standard used in email—organizations can publicly declare authorized agent identities, the certificate authorities (CAs) issuing them, and their allowed scopes.

Industry Implications

This shift toward standardized attestation is critical for highly regulated sectors, such as healthcare, where the ability to prove model integrity and maintain a tamper-evident trail of agent activity is a compliance necessity. By aligning its approach with NIST’s Cybersecurity Framework 2.0, IDC recommendations, and the IETF’s Workload Identity Management and Security Extensions (WIMSE), DigiCert is attempting to establish a cross-industry standard for AI governance.

As autonomous bots increasingly act without direct human oversight, the ability to cryptographically verify what an agent is and what it is allowed to do will be the primary defense against unauthorized data exfiltration and the bypassing of internal security controls.