A familiar voice used to feel like proof. It is not anymore.

In September, a North Texas family received an emergency call from a local number claiming their daughter had crashed her car and was in danger. The call appeared to use an AI-generated version of her voice. The family stopped and confirmed that she was safe before any money changed hands. The pattern is the warning: a familiar voice, a frightening story, and pressure to act before verifying.

The FBI’s latest U.S. data shows why this matters. In 2025, cyber-enabled crimes cost Americans nearly $21 billion. The agency received 22,364 complaints that referenced AI-related cybercrime, representing nearly $893 million in reported losses. That category covers several kinds of AI-enabled crime, not voice cloning alone. The FBI specifically notes that cloned voices can be used in business payment fraud and family “distress” scams.

Your ears are not an identity check

A scammer may find a short recording on social media, a podcast, a video, or a voicemail greeting and use it to imitate a family member, executive, client, or professional adviser. The call does not need to sound perfect. It only needs to sound convincing while you are worried, rushed, or trying to help.

The request usually creates pressure: send money now, keep this private, share a verification code, change payment instructions, or do not call anyone else. Those are not reasons to move faster. They are reasons to stop.

Recognizing the voice is no longer enough. Verify the person before you act on the request.

Agree on a code word before the call comes

Families should choose a private word or short phrase that can be used during an unexpected emergency call. Businesses can use a separate verification phrase for executives and employees who may make urgent requests by phone.

A useful code word should be

  • Unrelated: Do not use a birthday, address, pet, school, team, or anything visible online.
  • Memorable: Two or three unexpected words are easier to remember than a complicated string.
  • Private: Agree on it in person or during a call made to a number you already trust. Do not post it or place it in an obvious group-chat title.
  • Separate: Use different phrases for family emergencies and business requests.
  • Replaceable: Change it if anyone says it during a suspicious call or believes it may have been exposed.

Do not reuse a password, account PIN, security answer, or authentication code as the phrase. A code word is a checkpoint, not a credential.

A code word is not permission to move money

If the caller knows the phrase, continue verifying. Someone may have overheard it, found it in a compromised message, or pressured another person into sharing it.

For a family, hang up and call the person back using the number already saved in your phone. If you cannot reach them, contact another trusted person who can confirm where they are.

For a business, an urgent voice request should never replace the normal approval process. Payment changes, wire instructions, sensitive records, password resets, and account access should require a callback through a trusted company number, a written request in the system your team already uses, and a second authorized person when the risk is high.

The 90-second verification rule

  • Stop: Do not send money, disclose information, or share a login or authentication code.
  • Challenge: Ask for the private code word or phrase.
  • Disconnect: End the incoming call. Caller ID can be manipulated.
  • Call back: Use a number from your contacts, company directory, or another trusted source—not a number the caller provides.
  • Confirm: Verify the request with another family member, manager, or authorized approver.
  • Record: Document the attempt and alert anyone else who may be targeted.

Build the pause into the workflow

Training people to “be careful” is not a complete system. Put the verification steps where the request happens. A receptionist should know where to route a suspicious executive call. Finance should have a written rule for changed payment instructions. After-hours teams should know who can approve an exception and how to reach that person through a trusted channel.

AI can help flag unusual language, unexpected urgency, new payment details, or requests that fall outside a normal pattern. It should not be the only judge of whether a voice is real. The strongest defense is a repeatable process that gives a person permission to pause, verify, and refuse an unconfirmed request.

Where AI helps—and where IntelliLine steps in

AI is useful before and after the identity check. It can screen incoming communication for unusual urgency, compare a request with approved account information, create a record of what was said, and route the issue to the right person. It can make risk visible. It should not make the final decision on a voice alone.

IntelliLine Solutions turns those safeguards into a communication process your team can actually follow. We help businesses define which requests AI may handle, which ones require a person, how callers are verified, who approves sensitive actions, and what happens when a request arrives after hours. The result is not another tool sitting beside the workflow. It is a clearer path from the first call to a verified response.

What a stronger communication system should do

  • Recognize risk: Flag pressure, changed payment details, sensitive-data requests, and unusual caller behavior.
  • Pause the request: Prevent urgency from bypassing verification or approval.
  • Reach the right person: Route the issue to an authorized employee with the context needed to act.
  • Document the outcome: Record what was verified, who approved it, and what happened next.
  • Stay available: Apply the same rules during business hours, after hours, and when the owner is unavailable.

That is how AI supports better human interaction: it prepares the facts, catches warning signs, and protects the handoff so a trained person can use judgment when trust matters most. Our approach to operations and solutions are built around that balance.

My take

Voice cloning works by borrowing trust. The answer is not panic; it is preparation. Choose the code words now, establish the callback rule, and make sure everyone knows that urgency never cancels verification.

Sources