Entrust launches program to build trust controls for AI agents

Entrust has announced a co-development program aimed at helping enterprises put identity, authorization and cryptographic controls around autonomous AI agents. The company said the Agentic AI Trust Accelerator is intended to support organizations moving agentic AI projects beyond pilots and into production environments, particularly where compliance and auditability are material constraints.

The initiative comes as large organizations test AI agents that can act across applications, counterparties and internal processes with varying levels of human supervision. Entrust said enterprises need clearer ways to establish who approved an agent, what authority it has and how its activity can be verified after the event.

Entrust cited an IBM study that found 77% of CIOs and CISOs believe AI adoption is already running ahead of governance capabilities, while 59% identify security and compliance as leading obstacles to deployment. The company also pointed to Deloitte’s view that the move toward “human on the loop” automation requires stronger governance and oversight for AI agents.

Anudeep Parhar, Entrust’s chief operating officer for digital infrastructure and the executive leading the accelerator, said AI agents are developing faster than the controls needed to govern them. He said the program would bring customers and partners together to develop approaches for identity, authorization, cryptographic trust and accountability that can work with existing technology platforms.

How the trust layer is expected to work

Entrust said the accelerator will draw on its existing work in identity verification, cryptographic identity and lifecycle management for keys, certificates and secrets. Those tools are used to establish trusted digital identities, protect credentials and manage the cryptographic material that allows systems to authenticate, sign transactions and prove integrity.

For AI agents, the same concepts are being extended to autonomous actions. A verifiable identity can link a software agent to a human principal and to a distinct agent identity. Authorization controls can restrict what the agent may do under defined policies, roles and delegated authority. Cryptographic signing can create evidence that a specific agent performed a specific action at a given point in a controlled workflow.

Entrust said the program will focus first on regulated and compliance-driven settings, where organizations often need controls that operate across platforms, counterparties and business processes rather than inside a single application.

• Identity: verifiable human and agent identities that allow actions to be traced to a confirmed human principal and a unique agent.
• Authorization: real-time checks so agents operate within approved policies, roles and delegated authority, with human involvement for critical decisions.
• Cryptographic trust: protection for keys, certificates, secrets and signing functions used by agents to authenticate and transact.
• Accountability: cryptographically verifiable records for regulators, customers and internal risk teams.

Emanuel Figueroa, senior research analyst for identity security at IDC, said agentic AI shifts governance from a point-in-time exercise to a continuous discipline. He said organizations need continuous verification across the agent lifecycle as autonomous systems take on tasks associated with human workers.

Tony Ball, Entrust’s chief executive, said trust would influence how quickly organizations move from experimentation to production with agentic AI. Entrust said access to the accelerator will be limited to selected enterprise customers, financial institutions, cloud and software-as-a-service providers, systems integrators and other technology partners interested in developing trusted architectures for AI agents.

This story draws on original reporting from Finextra Research.