Identity Management: The Bottleneck for Scaling Enterprise AI Agents

As enterprises rush to deploy autonomous AI agents, a critical gap in operational control is emerging. As first reported by The Register, IBM's 2026 Cost of a Data Breach report indicates that 68% of organizations now lack the governance to manage or detect shadow AI, up from 63% the previous year. Furthermore, the percentage of organizations requiring IT approval for AI deployment has dropped from 45% to 38%.

DigiCert Senior Vice President of Product Brian Trzupek warns that many firms make the mistake of bolting agents onto existing human identity and access management (IAM) stacks using long-lived API keys. Trzupek notes that IAM was designed for humans who can manually approve requests, and relying on static credentials undoes the progress of zero trust architectures.

To solve this, DigiCert is promoting an "AI Trust" framework. Rather than extending human IAM, this approach aligns with recommendations from IDC, the IETF's Workload Identity Management and Security Extensions (WIMSE), and NIST's Cybersecurity Framework version 2.0 by framing agent identity as a workload identity issue. This strategy leverages the SPIFFE (Service Profile Identity) and SPIRE (SPIFFE Registration Endpoint) open standards to provide short-lived credentials and runtime attestation.

DigiCert also proposes using DNS as a verification point, similar to the DMARC email standard. By publishing agent policy records in DNS, organizations can declare authorized identities and allowed scopes. While IDC warns that scaling these records could create loopholes, Trzupek argues that automating the lifecycle and tying records to certificate expiration mitigates these risks. DigiCert's proposed architecture utilizes a single SPIRE server anchored to a DigiCert CA for identity, with a unified kill switch and policy enforcement via an Open Policy Agent engine.