The technology

An AI agent that guards identity in real time.

At the core of Vigil AI is an agent that coordinates five independent access controls. To break in, an attacker would have to fake a government-signed passport chip, a face, a live voice, a secret that never leaves the device and the physical device itself — all at the same moment.

Five controls

Independent layers, verified together.

01

Zero-knowledge password layer

The password never leaves the customer's device. Vigil AI verifies knowledge of it cryptographically, so there is nothing to phish, intercept or leak from a server.

02

NFC passport verification

The customer taps their passport to the phone. The chip inside is digitally signed by the issuing government and cannot be forged — a root of trust no deepfake can copy.

03

Facial recognition with liveness

Liveness detection tells a real person apart from photos, replays and synthetic deepfake video — and matches the face against the signed passport photo.

04

Changing-word voice verification

The customer speaks a word that changes every session — pre-recorded or cloned voices cannot anticipate it.

05

Bound Identity Token

A cryptographic token generated and stored in the device's Secure Enclave. It is non-transferable: it cannot be exported, copied or moved to another device.

Bound Identity Token

Identity anchored in hardware.

When a customer enrolls, Vigil AI binds their verified identity to a private key created inside the device's Secure Enclave — a hardware-isolated chip that even the phone's own operating system cannot read.

The key never leaves that chip, so the resulting token cannot be transferred, cloned or replayed from another device. Every sensitive action is signed by the enclave, giving the institution cryptographic proof that the real, enrolled customer is present.

Use case

A loan application, twice.

01

The customer applies

A real customer requests a loan. Passport chip, face, voice and device all match — approved in minutes.

02

The fraudster tries

An attacker submits the same request with stolen data, a deepfake video and a cloned voice.

03

Vigil AI blocks it

No signed passport chip, no enrolled Secure Enclave token. The agent flags the mismatch and the application is stopped.

The AI agent

Coordination is the moat.

Individual checks can each be attacked on their own. Vigil AI's agent evaluates every access request against all five controls together, weighs the signals in context, and adapts in real time when something looks wrong.

Because the controls are independent — something you know, something the government signed, something you are, something you say live, and something you hold — a single generative-AI technique can't defeat them all at once.

Architecture principles

Privacy-preserving by design

Customer data stays in the vault. Partners get time-limited access, not permanent copies.

Blockchain-backed records

Access grants and deletions are written to an immutable ledger that cannot be edited after the fact.

Built to integrate

Runs alongside an institution's existing fraud and identity systems instead of replacing them.

Building the privacy layer for AI-era finance.

Get in touch