Quick note before we start: I promised scheduled tasks this week. That moves to next week. Something more urgent came up, and it has to do with your phone.
On October 14, xAI (Elon Musk's AI company) released Grok Voice 3.0. The company says the same technology already answers about 20,000 Starlink support calls a day. It verifies who you are, troubleshoots your equipment and ships replacement hardware. It sounds natural, answers in under a second, and costs the company 8 cents a minute.
That last number is the one I can't stop thinking about. If a company can run a convincing voice for 8 cents a minute, so can a scammer.
I heard a version of this last month at an ACAMS conference. ACAMS is the Association of Certified Anti-Money Laundering Specialists, the people whose job is catching financial crime. In one wide-ranging panel discussion, the speed of fraud came up. Fraudsters and bad actors move fast. Companies move slower. New controls, new training and new rules all take time.
One point stuck with me: until the companies catch up, it's on us to educate the people we serve about the risks.
Put that next to 8 cents a minute. The tools just got cheaper for the people who move fastest.
So consider this issue my part of that job.
Nobody's paying me for this. No affiliate links. I don't use Starlink, and I'm not recommending or warning anyone off Grok. It's simply the clearest sign yet of how cheap and convincing these voices have become.
The calls coming in
The FBI's latest internet crime report covers 2025. Americans 60 and older reported losing $7.7 billion, about 60% more than the year before. Scams that involved AI accounted for $893 million of all losses, and $352 million of that came from people over 60. That includes cloned voices used in "family emergency" calls.
The FTC tells a similar story. People reported losing $3.5 billion to impostor scams in 2025, nearly triple the 2020 figure. Among business impostors, fake bank calls caused the biggest losses.
Those are only the losses people reported. Many never do.
Here's how the voice-clone version works. A scammer needs only a few seconds of someone's voice, often from a video posted online. Then Grandma gets a call from her "granddaughter": there's been an accident or an arrest, she needs money now, and please don't tell Mom. Last year an 86-year-old woman in Philadelphia lost $6,000 in cash this way. She said she never would have fallen for it if she hadn't recognized her granddaughter's voice.
Recognizing the voice used to be the safety check. That doesn't work anymore.
If you're watching out for a parent
A lot of us reading this are in the middle. We're old enough to be targets ourselves, and we're the ones looking after a mom or dad in their 70s or 80s. They're the ones who answer the house phone, the ones who were raised to be polite to whoever is calling, and the ones scammers go after first.
Four things worth doing this month, while nothing is wrong:
1. Pick a family safe word. Something only your family would know, never posted online. If "you" ever call in a panic asking for money, the first question is the safe word.
2. Make the rule out loud: hang up and call back. Any call about money, a bank account, Medicare or a family emergency gets hung up on and called back on a number they already have. Real banks and real grandchildren will understand.
3. Turn on call screening. Most carriers have a free scam-blocking tool (T-Mobile's Scam Shield and Verizon's Call Filter, for example). Fifteen minutes on their phone settings is worth it.
4. Ask their bank or brokerage about a "trusted contact." It lets the firm reach you if something looks off on your parent's account. It doesn't give you access to their money.
If your parent only remembers one of these, make it #2.
A safe word can be forgotten. Call screening misses things. A cloned voice can sound exactly like family. Hanging up and calling back on a number you already trust works every time, no matter how good the voice on the other end is.
And it helps to know this: as a rule, banks and other financial institutions don't call you and ask for your information. They already have it. If a call asks you to confirm an account number, a password, a PIN or a code that was just texted to you, that's your signal to hang up. Then call the number on the back of your card and ask them yourself.
The rules are starting to catch up (in some places)
Since 2024, the FCC has treated AI-generated voices in robocalls the same as any other recorded voice. Without your prior consent, they're illegal. That helps against legitimate companies cutting corners. It doesn't stop a scammer who's already breaking the law.
On the legitimate side, California signed AB 1609, the "Right to Human Customer Service Act," on September 28. It takes effect January 1, 2027, and applies to companies with more than $500 million in revenue that serve Californians:
• A company can't pass its chatbot off as a human. If you could reasonably be fooled, it has to tell you it's AI.
• You must have a simple way to ask for a person, at least during business hours.
• The company has to make a good-faith effort to connect you within 15 minutes, or schedule a callback within one business day.
Here in South Carolina, we don't have anything like it. But when a state as big as California sets a rule, large companies often adopt it nationwide rather than run two systems. That's a likelihood, not a promise.
AARP's Senior Planet released research on October 1: 70% of older adults said they aren't confident about what information is safe to share with AI tools.
That worry makes sense, and it matters most on a phone call. The rule hasn't changed just because the voice might be AI. If you called a number you trust, the company can verify you with basic details. You never read out a full Social Security number, a full card number or a password because a voice asked for it, human or not.
Try this
On a call you placed:
1. Ask directly: "Am I talking to a person or an automated assistant?" A legitimate company's system should give you a straight answer.
2. Ask for a person by name: "Representative" or "speak to a person" are the phrases these systems usually recognize. If the first try doesn't work, say it again.
On a call you didn't place:
3. Urgency is the tell. Pressure to act now, keep it secret, or pay with gift cards, wire transfers or crypto means hang up. No real bank or grandchild needs that.
4. Hang up and call back on the number printed on the card, the statement, or in your own contacts. That rule is older than AI. It matters more now.
You've spent decades reading people on the phone. That instinct still counts, but a familiar voice isn't proof anymore. A callback is.
Your turn
Have you or a parent gotten a call this year that didn't feel right? What tipped you off?
Hit reply and tell me. I read every one.
Next week: scheduled tasks, as promised.
— Eric
Connect with me on LinkedIn: linkedin.com/in/eric-skelton-ai-over-50-newsletter