Share with your CISO
Entrust is betting that the governance gap around autonomous AI agents is large enough to build a business around, launching its Agentic AI Trust Accelerator as a co-development programme with enterprise customers and technology partners. The programme targets four control layers: verifiable identity for agents and humans, real-time authorisation, cryptographic key and certificate management, and proof-of-action audit records. An IBM study Entrust cites puts the urgency in numbers: 77% of CIOs and CISOs say AI deployment is outpacing governance, and 59% flag security and compliance as active blockers. Initial intake is limited to financial institutions, cloud providers, and systems integrators.
What this means for your business
The question this programme forces isn’t whether your organisation needs agent governance, it’s whether you have enough of it already to stay out of a co-development arrangement that is, by design, still figuring things out. If your AI agents touch regulated workflows, cross organisational boundaries, or operate with delegated authority over financial or HR systems, the four control layers Entrust is building around aren’t optional features. They’re the audit trail regulators will ask for when something goes wrong. Companies already deep in agentic deployments without those controls are the obvious target customer here, and they know who they are.
The structural problem Entrust is addressing has a name inside security architecture circles: the trust plane, the layer that sits beneath model behaviour and above raw infrastructure, governing what an agent is allowed to claim, do, and prove it did. Today that layer barely exists in most enterprises. Platform-level policies from model vendors handle some of it. Identity providers handle pieces. PKI (public key infrastructure, the system that issues and verifies digital credentials) handles cryptographic trust for humans and machines but wasn’t designed with autonomous software agents in mind. Entrust’s existing position in PKI and certificate management gives it a credible entry point, though it’s building from adjacent territory rather than from a standing start in agent-native tooling.
The accelerator format is worth reading carefully. Entrust isn’t shipping a product; it’s co-developing reference architectures with paying enterprises, which means early participants are funding the R&D while getting first access to the outputs. That’s a reasonable trade if you’re a large financial institution that needs a governance framework regardless, and you’d rather shape the standard than inherit someone else’s. It’s a worse trade if you’re looking for a tested, deployable solution now. If your agentic AI rollout is measured in months, not quarters, this programme won’t close your governance gap on your timeline. I’d revise that assessment if Entrust announces a generally available product release within the next two fiscal quarters rather than continuing to grow the co-development cohort.
Concept deep-dive: Proof-of-action records
A proof-of-action record is a cryptographically signed log that captures what an AI agent did, when it did it, and under whose authority. Think of it as the equivalent of a notarised transaction receipt, except generated automatically for every action an agent takes in a business system. Regulators in finance and healthcare already expect this kind of audit trail for human operators. Extending it to software agents is the same requirement applied to a new class of actor, and it’s the piece most current AI platforms don’t produce natively.
Based on reporting from Entrust launches AI trust accelerator for autonomous agents, originally published 2026-07-15 03:30:00.
