Entrust announced the Agentic AI Trust Accelerator on July 14, a co-development program meant to give enterprises the identity plumbing autonomous agents currently lack. The pitch is straightforward: companies are running AI agents against real systems but can't answer basic questions once something goes wrong — who authorized the agent, what it was permitted to do, and how its actions can be proven after the fact. The program brings together a limited set of enterprise customers, financial institutions, cloud and software providers, and systems integrators to build what Entrust calls a "trust plane" for autonomous AI, built on verifiable identity, real-time authorization and cryptographic proof of action that can travel across systems, partners and workflows. It's structured around four pillars: identity, authorization, cryptographic assurance, and accountability. The identity component is meant to verify both human users and AI agents while ensuring every agent action can be traced back to a responsible individual, and authorization is meant to limit agents to approved roles, policies, and permissions, with human oversight built in when needed. This isn't a green-field build. Entrust is drawing on its existing PKI, certificate, and secrets-management business — the same infrastructure that already underwrites TLS certs and identity verification for human users — and extending it to agent workloads. Entrust COO Anudeep Parhar framed the problem as a pacing mismatch: AI agents are advancing faster than the trust infrastructure needed to govern them. CEO Tony Ball's framing, that trust will determine how quickly companies can move agentic AI from experimentation to production, points at the actual bottleneck: it isn't model capability, it's the absence of a chain-of-custody layer that regulated industries can point to during an audit. Coverage of the launch notes the program is explicitly aimed at regulated and compliance-driven sectors, and one recap ties it to a specific pain point: 77% of CIOs report that AI adoption is outpacing their governance capabilities. That figure should be read as a vendor-cited stat rather than independently verified, but it matches the shape of a problem several other outfits are converging on this year — NIST's AI Agent Standards Initiative, Microsoft's Entra Agent ID rollout, and the DIF's stewardship of the MCP-I identity framework all describe the same gap from different angles: agents need identities that are cryptographically bound, not borrowed from a human's login session. What makes this one worth flagging over the others is that it's explicitly a co-development effort rather than a shipped product — Entrust is opening participation to a small cohort of enterprises and partners to validate use cases before anything hardens into a general offering. That's a meaningful signal about where this category actually is: still assembling reference architectures, not yet standardizing them. Whether Entrust's four-pillar framing becomes a template others adopt, or just one more vendor's take on a problem several parties are racing to solve, will depend on whether the accelerator produces interoperable specs or a closed pipeline that only works inside Entrust's own PKI stack.