AI ToolsA serious build1 evidence point

A “Prove You’re Human” Layer

Let apps and video calls confirm the person on the other end is real before trusting them.

$50–$500 per customer, per month

Overview

What this is

A verification layer other software calls before it trusts a person — a check that the account, the caller or the face on the video call is a live human and the one they claim to be, not a synthetic stand-in. Delivered as an API and a drop-in widget rather than something an end-user installs, so a bank, a marketplace or a support tool can put a real-human gate in front of the moments that matter: a high-value transfer, a new account, an "approve this" call.

The problem it solves

Almost every online system quietly assumes the entity on the other end is a real, consistent person — and generative audio and video have made that assumption unsafe. Deepfaked colleagues and customers are now good enough to move money and pass live calls; a single faked video call has already cost one company around $25M. The old signals — a face, a voice, a video feed — are exactly the ones that can now be faked cheaply.

Why now

The threat crossed from theoretical to operational fast, and the businesses exposed to it — finance, marketplaces, anything with a support line that can be socially engineered — are actively looking for a defence right now. Y Combinator named it a Fall 2026 Request for Startups: a verification layer banks, apps and video calls run before trusting anyone.

The evidence

  • Y Combinator's Fall 2026 Requests for Startups asks for "Proving You're Human" — a verification layer banks, apps and video calls check before trusting anyone, citing that deepfaked colleagues already cost one company $25M on a single call.

    Y Combinator — Requests for Startups (Fall 2026) →

What to watch out for

This is an adversarial security product: the people you are defending against improve every month, so a check that works today can be beaten tomorrow, and a false sense of security is worse than none. You are also handling biometric and identity data, which carries real privacy law and real breach consequences — treat both accuracy and data handling as existential, not as polish.

A sensible starting stack

Liveness / anti-spoof models, An identity data provider, A low-latency API, Any web stack

Keywords: identity verification, deepfake detection, liveness, fraud prevention, kyc, trust and safety