iProov has warned that the rapid deployment of autonomous AI agents in enterprise environments is creating an “accountability vacuum” in which high-impact decisions are made without verifiable human authorization, and has demonstrated cryptographic methods for binding AI agent actions to confirmed human intent at RSA Conference 2026.
The company argues that authentication standards currently in widespread use, including FIDO2, one-time passcodes, and push notifications, were designed with a human at the keyboard. These systems verify identity and permission but cannot verify intent. As AI agents increasingly initiate transactions and access systems autonomously, that assumption breaks down. iProov frames the gap as a problem that is structurally similar to the deepfake-driven erosion of public trust, but generated from within organizations’ own automated systems.
iProov’s founder and CEO Andrew Bud has outlined three requirements for meaningful oversight: the right human with the authority to make the decision, sufficient context for that person to make a real choice, and an attributable, timestamped record tied to a verified identity. Without these elements, the company argues, organizations risk legal, financial, and ethical exposure as AI-agent-driven decisions accumulate across operations.
At RSA 2026 in San Francisco, iProov’s Johan Sellström demonstrated how AI agent actions can be cryptographically tied to verified human approval, producing an auditable record that links a decision to an identified person at a specific time. The company says this kind of binding is necessary for organizations to scale agentic AI while maintaining defensible oversight.
iProov launched a workforce identity suite earlier this year focused on verification scenarios beyond standard login, positioning biometric checks as infrastructure for higher-stakes enterprise processes. The accountability vacuum argument extends that positioning to cover AI agent authorization specifically.
The US National Cyber Strategy calls for rapid adoption of agentic AI to modernize national-scale systems, while NIST has published a concept paper identifying human-in-the-loop binding as a core requirement for agent identity frameworks. NIST has also been developing guidance on AI agent identity controls, and vendors including Wink and Vouched have begun integrating biometric proof of personhood into AI agent workflows. iProov published its research on the topic under the title “The Great Trust Recession.”
Sources: iProov, iProov Blog
–
By the ID Tech Editorial Team
