Scale with AI / Episode 12 / October 15, 2026
Scale with AI — October 15, 2026
AI voice clones and deepfakes are now aimed straight at small businesses — one fake video call cost a company more than twenty-five million dollars. This episode covers the three-rule protocol that stops payment scams cold, Pennsylvania's new AI-calling consent law taking effect in three days, and Maryland's new deepfake fraud law.

Quick Take
- If a voice on the phone asks for money, verify before you pay — a cloned voice can't pass a callback and a code word.
- The protocol is three rules: call back on a number you already trust, keep a code word only your real team knows, require two-person approval above a set amount.
- Pennsylvania Act 47 takes effect October 18 — documented written consent for automated and AI-voice calls, 9AM–7PM Monday–Saturday, no Sundays or holidays.
- Maryland's deepfake fraud law is now in effect: AI impersonation for fraud is a felony.
The Shift
The scam has changed. The FBI counted more than twenty-two thousand AI-related fraud complaints in twenty twenty-five, with losses approaching nine hundred million dollars, and IBM's latest breach report finds one in four malicious breaches now involves AI — deepfake impersonation chief among them. At the same time, most small businesses are adopting AI fast and loosely: in Pax8's Q3 survey of over a thousand SMBs, sixty-seven percent of AI-using businesses had built their own solution, and thirty-five percent discovered employees running unsanctioned 'shadow AI' in the past six months. The opportunity and the exposure are growing on the same curve.
The Fine Print
Two laws matter this month. Pennsylvania Act 47 takes effect Sunday, October 18: automated and AI-voice calls require documented prior express written consent, calls are limited to nine AM to seven PM Monday through Saturday with no Sundays or holidays, and if you have no consent record, the law treats consent as absent. Maryland's deepfake fraud law took effect October 1, making AI-impersonation fraud a felony. The through-line: regulators now treat synthetic voices as a serious fraud vector — and they expect businesses to keep proof.
Who’s Doing It
The businesses beating voice-clone scams aren't buying exotic technology — they're running a protocol. When a payment request arrives by phone or video, they hang up and call back on a number already in their records. They keep a rotating code word that only the real team knows, so a cloned voice can't pass. And any payment above a set threshold needs a second human's approval. Those three habits would have stopped the most famous case on record: a finance worker paid out more than twenty-five million dollars after a video call where every other 'executive' on screen was a deepfake. On-device fake-call detection is also starting to arrive on phones — helpful backup, never the plan itself.
Your Move
This week, install the three rules: callback on a known number, a code word, and two-person approval over your threshold — and tell your team a real executive will never punish them for verifying. Then run the compliance side before Sunday: confirm every automated or AI-voice call you make has documented consent behind it and respects Pennsylvania's new hours. Not sure where your business leaks time, money, or risk? Start with The Agency's free AI audit — link in the show notes — or book a conversation and we'll map it with you.
Key Takeaways
- AI fraud is now a line item: ~22,000 FBI complaints and nearly $900M in losses in 2025, and 1 in 4 malicious breaches involves AI (IBM).
- A three-rule protocol — callback on a known number, a code word, two-person approval — defeats voice-clone payment scams without new software.
- Pennsylvania Act 47 takes effect October 18: documented written consent for AI/automated calls, 9AM–7PM Mon–Sat, no Sundays or holidays.
- Maryland's deepfake fraud law is now in effect — AI impersonation for fraud is a felony.
Full transcript
Alex: Welcome to Scale with A I for Thursday, October fifteenth, twenty twenty-six. I am Alex.
Jordan: And I am Jordan. Today, the other side of the A I boom. The same technology that can answer your phone and book your jobs can also copy a voice in seconds and use it to try to steal from your business. The good news is the defense is simple, and the businesses putting it in place now are sleeping better than the ones hoping it will not happen to them.
Alex: Let us start with The Shift. A I has moved from something your business tries out to something your business quietly runs on. A new survey of more than one thousand small business leaders, released this week, found that two thirds of businesses using A I have now built a tool of their own in regular daily use. Not a trial. A working part of the operation.
Jordan: And here is the twist. About one in three discovered A I running somewhere in the company that nobody had approved. A team member connected a free tool to the shared files to draft quotes faster. Someone automated a follow up that touches customer information. Nobody meant harm. Nobody checked either. It just runs.
Alex: That is the shift in one sentence. The question is no longer whether A I is inside your business. It already is. The question is whether you chose it and secured it, or whether it arrived by accident and stayed by default.
Jordan: Meanwhile, pressure is arriving from outside. Voices can now be cloned from a very short clip. A voicemail greeting. A social video. A few seconds is enough to sound like an owner, a manager, or a family member in distress.
Alex: The losses are not hypothetical. Federal investigators counted tens of thousands of A I related complaints in the most recent full year, with reported losses in the hundreds of millions of dollars. A separate breach study this year found about one in four malicious breaches now involves A I, and impersonation is the most common trick. In the most famous documented case, a finance worker joined a video call where every other face was fake, and sent out tens of millions of dollars before anyone knew.
Jordan: So how does the scam work, mechanically?
Alex: It needs three ingredients every time. First, authority. The request sounds like it comes from someone you obey. Second, urgency. It has to happen right now. Third, one channel. The whole thing happens inside a single call or meeting, so you never get an independent chance to check. Remove any one of those three, and the scam falls apart.
Jordan: That leads straight into The Law.
Alex: Start with Pennsylvania, because the clock is real. In three days, on October eighteenth, Pennsylvania's updated telemarketing law takes full effect. If you call or text customers there, the window shrinks to nine A M to seven P M, Monday through Saturday. No Sundays. No legal holidays. And it is not just calls. Texts, voicemails, and ringless voicemails all count.
Jordan: The consent bar rises too.
Alex: Automated and A I voice marketing now needs prior express written consent, a signed agreement that names the number, says clearly what the person is agreeing to, and says agreeing is not a condition of buying anything. If a customer replies stop or unsubscribe, that stop has to work. No record of consent means, in practice, no consent.
Jordan: And Pennsylvania is not alone. A new state law that took effect at the start of this month makes it a felony to knowingly use A I to impersonate a real person to commit fraud, with civil claims on top. The direction is consistent. Consented, disclosed, verifiable communication is becoming the standard.
Alex: Which brings us to Who Is Doing It, the owners beating the scam with process instead of panic.
Jordan: What do the prepared businesses actually do?
Alex: They run a three rule protocol, and every employee who can move money knows it is not optional. Rule one, call back. Any request to pay, change bank details, or reset a password that arrives by voice or video gets verified by hanging up and calling a number already on file. Never a number supplied inside the request.
Jordan: Rule two?
Alex: A code word. People who authorize sensitive work share a rotating word. The real owner knows the current one. A cloned voice does not. It takes ten minutes to set up and defeats a convincing fake in seconds.
Jordan: And rule three caps the damage.
Alex: Two person approval above a threshold you choose. No single person, acting on a single call, can move serious money alone. Notice what these rules share. None asks your team to spot a fake by ear. Detection is improving, and new phones can now flag a suspected fake right on the device. That helps. But winning businesses do not bet the bank account on anyone's ear. They design the process so a perfect fake still cannot get paid.
Jordan: They also do the inside half. A quiet audit. Which A I tools is the team actually using, what information flows into them, and who approved them. A found tool can be secured or replaced. An unfound one just keeps running.
Alex: So, Your Move. Three moves, all finishable before the weekend. First, set the protocol. Pick your payment threshold, name the second approver, choose the code word, and tell the team the callback rule is absolute. Second, run the shadow check. Ask every team member, without blame, what A I tools they use for work, and write them down. Third, run the consent and hours check on your own outreach before the Pennsylvania deadline on Saturday. Are consent records findable, are calling hours set to the customer's time zone, and does stop actually stop everything?
Jordan: And if that list comes back messier than you expected, that is exactly the work we do at The Agency. We build the systems that answer, book, and follow up for businesses like yours, and we build the consent, verification, and approval guardrails around them, so the technology that saves you time never becomes the technology that costs you. Start with our free A I audit at theagencyllc dot net, and see exactly where your business stands.
Alex: I am Alex.
Jordan: And I am Jordan. Verify the request, not the voice, and we will see you tomorrow on Scale with A I.
Sources
Your next conversation