Token has extended its biometric authentication technology to enterprise AI agents, positioning a live human fingerprint check as a mandatory gate before an agent can carry out a high-consequence action. The approach treats biometric approval as a deterministic control point that sits outside the agent’s own reasoning, so that actions like moving money or deleting data require a verified person to sign off rather than relying on the agent’s judgment alone.
The control is built on Token’s biometric hardware. The company’s devices require a live fingerprint match against credentials held in secure hardware, bound to the user and to physical proximity, without passwords, one-time codes, push approvals, or cloud-synced secrets that can be phished or replayed. Applied to AI agents, that hardware becomes what Token calls a hard gate: before an agent executes a sensitive operation, it must obtain a biometric approval from an authorized human.
The gate is aimed at the actions where an autonomous agent can do real damage. Token lists financial transactions and vendor payments, data deletion, changes to access rights, release of confidential information, and modifications to production systems among the high-consequence operations that would sit behind a biometric check. The design is meant to hold even if an agent is hijacked or behaves unexpectedly, because the approval step is a separate control rather than another software guardrail the agent could be manipulated into bypassing.
“AI agents are becoming part of the enterprise operating system. That is incredibly powerful, but it also means agents need real control points,” said Kevin Surace, Chief Executive Officer of Token, contrasting biometric approval, which he called deterministic, with monitoring approaches that remain probabilistic. The pitch is that speed of automation is preserved for routine actions while the riskiest ones are stopped pending a human fingerprint.
The move extends Token’s biometric hardware line into the agentic-AI market. The company built its enterprise business on biometric security keys with a fingerprint sensor and a secure element, and is now applying that hardware-rooted model to the problem other vendors are addressing in software, as with Anonybit’s platform for AI agent identity security.
Sources: SourceSecurity
–
By the ID Tech Editorial Team
