DigiCert Proposes AI Trust Framework to Combat Shadow AI and Governance Gaps
DigiCert has introduced an AI Trust framework leveraging PKI, DNS, and attestation to address the growing problem of shadow AI and governance deficits in enterprise environments.
The rapid proliferation of artificial intelligence agents within enterprise environments is creating a significant governance challenge, often referred to as 'shadow AI.' These autonomous and adaptable agents, while powerful, introduce risks that traditional security models struggle to address. Brian Trzupek, senior vice president of product at DigiCert, notes a familiar pattern: the allure of advanced technology often leads organizations to bypass security protocols in their haste to adopt it. This trend is reflected in IBM's 2026 Cost of a Data Breach report, which indicates a rise in organizations lacking adequate governance for AI and a decrease in those requiring IT approval for AI deployments.
To combat this growing threat, DigiCert has developed its AI Trust framework. This approach builds upon DigiCert's core competencies in public key infrastructure (PKI), DNS, and attestation. The framework aims to provide enterprises with the tools to answer critical AI governance questions, including identifying the AI agents in use, tracking regulated data flow to these agents, understanding the credentials they possess, enabling immediate containment of compromised agents, and ensuring the reconstructability of security incidents through tamper-evident trails.
A primary hurdle for many organizations is the sheer difficulty in identifying and tracking AI agents. Trzupek highlights a common mistake: attempting to shoehorn AI agent management into existing human identity and access management (IAM) systems. This approach often leads to the use of static, long-lived API keys with excessive permissions, undermining zero-trust principles. Instead, industry bodies like IDC and standards organizations such as NIST and IETF are advocating for treating AI agent identity as a workload identity problem, separate from human IAM.
This shift towards workload identity aligns with emerging standards like SPIFFE (Secure Production Identity Framework for Everyone) and its implementation, SPIRE. These standards enable agents to be managed as governed workloads that require runtime attestation and short-lived credentials, a more secure and scalable approach than traditional IAM methods. DigiCert's strategy incorporates these principles, aiming to provide a unified identity solution for diverse AI agents.
DigiCert's framework also proposes leveraging DNS as a critical governance tool. By publishing agent policy records in DNS, organizations can declare authorized agent identities, the certificate authorities that issued their credentials, and their permitted operational scopes. Gateways can then query these DNS records to verify agent legitimacy, terminating sessions for unauthorized agents. This approach mirrors the DMARC standard used for email security, providing a centralized point for policy enforcement.
While the scale of managing DNS records for a potentially vast number of AI agents presents a challenge, DigiCert argues that DNS has historically scaled to meet similar demands. Automating the lifecycle of these records, tying them to certificate issuance and expiration, can mitigate these concerns. The company emphasizes that overly permissive scope declarations are more of a discipline issue for operators than an architectural flaw.
DigiCert's proposed solution, dubbed the AI Agent Passport, acts as a cryptographically protected artifact that links an agent's identity to its authorization. This passport contains essential information about approved systems and permitted operations, ensuring that even third-party agents like Microsoft Copilot or Salesforce Agentforce can be managed within a consistent governance framework. The ultimate goal is to provide a unified control plane for AI agent security, regardless of their origin or function.