RefugeIDby NausicaAI

About

RefugeID is call screening for the AI era.

Software that answers first and tells your family whether the voice on the other end is really someone they know, before anyone reaches for the phone.

What RefugeID does

Every unknown call is screened before a phone in your household rings. RefugeID analyses the audio from the first moments of a call for the artifacts that AI-generated and cloned speech leave behind, matches the pattern against a library of known scam scripts, and compares the caller to an approved list you control. Clean calls ring through normally. A call that looks synthesized is held instead, labelled clearly on screen, and sent to the family guardian as an alert with the number and what was detected. No recordings are kept: the audio is analysed to answer one question and then dropped.

Who it is for

Households where the phone is a risk rather than a convenience. Aging parents who answer every call and are targeted by grandparent scams, fake bank fraud and government impersonation. Children with a first phone number, whose unknown callers we screen by default. And the adult children who manage the family's phones and want to know the moment something is wrong: one account, one guardian, every protected line visible in one place. RefugeID is $8 per month per protected line, and families who pre-order take 25% off their first year when they pay for the year up front.

Patterns we screen for

The scripts scammers reuse, described plainly. Each one is documented by a regulator, a law-enforcement report or a published case, and every description links to the source.

Cloned-voice family emergency

critical

A caller uses a few seconds of audio, often lifted from social media, to sound like a child or grandchild in trouble. The call opens mid-emergency, keeps the listener from checking with anyone else, and pushes for a payment that cannot be reversed. In every documented case the family member was safe and the demand was extortion.

  • An unfamiliar number opens with a relative's voice already in distress
  • A demand for money that cannot wait and cannot be verified
  • A request not to call the relative back or tell anyone else
  • Payment by gift card, wire transfer or crypto only
Federal Trade Commission consumer alert

Deepfake-authorised payment request

critical

A synthesized voice of a boss, colleague or bank official authorises a transfer. Documented cases have defeated payment controls because the caller sounded exactly like the person with authority to approve it. The technique is the second most severe category in the AI Safety Research Group register, with a single documented loss of $25.6M.

  • A payment request that arrives by phone with no written confirmation
  • Urgency tied to a closing window or an account about to be frozen
  • Instructions to keep the transfer off the usual approval path
AI Safety Research Group risk register

Official impersonation call

high

Callers pose as police, tax authorities or a utility, sometimes with a synthetic version of an official's voice, and threaten arrest, fines or a shut-off unless a payment is made during the call. Authority is spoofed at the scale of a phone bank, with enforcement arriving only afterwards.

  • Threats of arrest or account closure within minutes
  • Payment demanded over the phone while you are still on the call
  • A request for a one-time code, password or account PIN
  • A caller insisting you stay on the line while you fetch money
AI Safety Research Group risk register

Voiceprint treated as a password

watch

Bank and support lines that verify identity by voice can be attacked with synthesized speech. The current research benchmark carries dozens of attack algorithms aimed at speaker verification, so a matched voiceprint should never be the only check on an account reset.

  • An account reset that relies on your voice alone
  • A support line that accepts a few spoken words as proof of identity
  • A request to record you repeating a phrase for verification
AI Safety Research Group risk register

Who builds it

NausicaAI

RefugeID is built by NausicaAI, a small independent product team based in the United States. We are not a carrier, a data broker or an advertising company, so the product is a subscription for families and nothing else. That is what lets us say plainly that call audio is never sold, never used to train third-party models and never stored. Our engineering follows published evidence rather than headlines, drawn from the AI Safety Research Group's risk register and the regulators, law-enforcement reports and technical research behind it. Every figure we cite links to the original document.

Read the research and sourcesContact the team
RefugeID

Know who is calling before anyone picks up. RefugeID is AI voice detection and call screening that flags cloned voices and phone scams for the families who need it most.

Company

  • NausicaAI
  • Founded for families
  • United States

© 2026 NausicaAI. RefugeID is a NausicaAI company.

No recordings kept. No voices sold.