Can You Still Trust a Video Call?
Three stories this week. One is a safety warning worth passing along to your family, one is a change to a free tool we’ve recommended, and one is a handy new feature for iPhone users. As always, we’ll keep it plain and skip the hype.
1. An AI That Can Hold a Video Call Fooled Nearly Half Its Testers
On October 1, an AI company called Tavus showed off a new system named Griffin. It can carry on a live video call. It sees you, hears you, and answers back with a realistic face and voice as the conversation happens.
In the company’s own test, 54 people had a one-minute video call without being told whether a person or a computer was on the other end. Afterward, 26 of them said they thought they’d been talking to a real person. That’s almost half.
Before anyone panics, here are some caveats. Tavus ran the test itself, the calls lasted only a minute, and nobody outside the company has checked the results. Tavus also says it isn’t releasing Griffin to customers yet, because it knows the same realism could be used to fool people. So scammers don’t have this particular tool.
Why it matters to you: For years, a common piece of advice has been “If you’re not sure who’s calling, ask them to video chat.” That advice doesn’t hold up the way it used to. Scammers already use AI to copy voices, and convincing fake video is the obvious next step. A familiar face on a screen doesn’t prove anything by itself anymore.
The protections that work best don’t rely on what you see or hear, so a better fake doesn’t beat them:
- Hang up and call back. Use a number you already have for that person, not one the caller gives you.
- Use a family code word. Pick a word or phrase only your family knows. If someone claiming to be a loved one can’t say it, you have your answer.
- Ask something only they would know. “What did we have for dinner last Sunday?” is hard for a stranger to fake.
- Slow down when money comes up. Any call, voice or video, that asks for money, gift cards, a wire transfer, or secrecy (“don’t tell Mom”) is a red flag.
For the full rundown, read: Is AI Safe for Seniors? What You Need to Know →
2. Free Google Gemini Is Changing on October 9
Starting Friday, October 9, people who use the Gemini app without paying will get only Google’s smallest, lightest model, called Flash-Lite. The more capable versions will need a paid plan:
- Free: the lightweight model only
- Google AI Plus (about $5 a month): adds Google’s mid-level model
- Google AI Pro ($19.99 a month): includes everything
Free Gemini isn’t going away. The lighter model handles quick, everyday questions just fine, like a recipe idea, what a word means, or a fast fact. You may notice a difference on longer or trickier requests, where answers could come back simpler or less thorough.
You don’t need to do anything, and please don’t feel pressured to upgrade. If free Gemini ever gives you a thin answer, ask the same question in the free version of ChatGPT or Claude. Comparing two answers is a good habit anyway.
Not sure which AI tool fits you best? See: 7 Best AI Tools for Seniors in 2026 →
3. ChatGPT Can Now Scan Multi-Page Letters (iPhone First)
The camera in the ChatGPT app on iPhone has a new Scan option. You photograph several pages in a row, and ChatGPT stitches them into one document you can ask questions about.
This is handy for confusing multi-page mail, like an insurance explanation of benefits, a notice from the utility company, or the fine print on an appliance warranty.
To try it:
- Open the ChatGPT app on your iPhone.
- Tap the camera icon.
- Tap the three-dot menu and choose Scan.
- Photograph each page, one after another.
- Ask something like: “Explain this letter in plain English. Is there anything I need to do, and by when?”
It’s rolling out gradually, so if you don’t see Scan yet, give it a few days. There’s no Android date yet.
Before you scan, cover anything sensitive, such as your Social Security number, Medicare number, or bank account numbers. A sticky note works fine. ChatGPT doesn’t need those numbers to explain what a letter means.
New to ChatGPT? Start here: ChatGPT for Seniors: A Plain English Beginner’s Guide →
Your One Small Action This Week: Pick a Family Code Word
It takes about five minutes. Bring it up at dinner or on your next call with the kids and grandkids. Choose a word or short phrase that isn’t anywhere online. That means it shouldn’t be a pet’s name, your street, or anything you’ve posted on Facebook. Then agree on one rule: if anyone ever calls or video chats asking for emergency money, they get asked for the word.
It costs nothing. And because the word was never written down anywhere online, there’s nothing for a scammer’s AI to find.
Know someone who should see this? Share this article with your family, especially the ones who’d be first to send help if they thought you were in trouble. The code word only works if everyone knows it.
This Week in AI, October 5, 2026: Can You Still Trust a Video Call?
Three stories this week. One is a safety warning worth passing along to your family, one is a change to a free tool we’ve recommended, and one is a handy new feature for iPhone users. As always, we’ll keep it plain and skip the hype.
1. An AI That Can Hold a Video Call Fooled Nearly Half Its Testers
On October 1, an AI company called Tavus showed off a new system named Griffin. It can carry on a live video call. It sees you, hears you, and answers back with a realistic face and voice as the conversation happens.
In the company’s own test, 54 people had a one-minute video call without being told whether a person or a computer was on the other end. Afterward, 26 of them said they thought they’d been talking to a real person. That’s almost half.
Before anyone panics, here are some caveats. Tavus ran the test itself, the calls lasted only a minute, and nobody outside the company has checked the results. Tavus also says it isn’t releasing Griffin to customers yet, because it knows the same realism could be used to fool people. So scammers don’t have this particular tool.
Why it matters to you: For years, a common piece of advice has been “If you’re not sure who’s calling, ask them to video chat.” That advice doesn’t hold up the way it used to. Scammers already use AI to copy voices, and convincing fake video is the obvious next step. A familiar face on a screen doesn’t prove anything by itself anymore.
The protections that work best don’t rely on what you see or hear, so a better fake doesn’t beat them:
- Hang up and call back. Use a number you already have for that person, not one the caller gives you.
- Use a family code word. Pick a word or phrase only your family knows. If someone claiming to be a loved one can’t say it, you have your answer.
- Ask something only they would know. “What did we have for dinner last Sunday?” is hard for a stranger to fake.
- Slow down when money comes up. Any call, voice or video, that asks for money, gift cards, a wire transfer, or secrecy (“don’t tell Mom”) is a red flag.
For the full rundown, read: Is AI Safe for Seniors? What You Need to Know →
2. Free Google Gemini Is Changing on October 9
Starting Friday, October 9, people who use the Gemini app without paying will get only Google’s smallest, lightest model, called Flash-Lite. The more capable versions will need a paid plan:
- Free: the lightweight model only
- Google AI Plus (about $5 a month): adds Google’s mid-level model
- Google AI Pro ($19.99 a month): includes everything
Free Gemini isn’t going away. The lighter model handles quick, everyday questions just fine, like a recipe idea, what a word means, or a fast fact. You may notice a difference on longer or trickier requests, where answers could come back simpler or less thorough.
You don’t need to do anything, and please don’t feel pressured to upgrade. If free Gemini ever gives you a thin answer, ask the same question in the free version of ChatGPT or Claude. Comparing two answers is a good habit anyway.
Not sure which AI tool fits you best? See: 7 Best AI Tools for Seniors in 2026 →
3. ChatGPT Can Now Scan Multi-Page Letters (iPhone First)
The camera in the ChatGPT app on iPhone has a new Scan option. You photograph several pages in a row, and ChatGPT stitches them into one document you can ask questions about.
This is handy for confusing multi-page mail, like an insurance explanation of benefits, a notice from the utility company, or the fine print on an appliance warranty.
To try it:
- Open the ChatGPT app on your iPhone.
- Tap the camera icon.
- Tap the three-dot menu and choose Scan.
- Photograph each page, one after another.
- Ask something like: “Explain this letter in plain English. Is there anything I need to do, and by when?”
It’s rolling out gradually, so if you don’t see Scan yet, give it a few days. There’s no Android date yet.
Before you scan, cover anything sensitive, such as your Social Security number, Medicare number, or bank account numbers. A sticky note works fine. ChatGPT doesn’t need those numbers to explain what a letter means.
New to ChatGPT? Start here: ChatGPT for Seniors: A Plain English Beginner’s Guide →
Your One Small Action This Week: Pick a Family Code Word
It takes about five minutes. Bring it up at dinner or on your next call with the kids and grandkids. Choose a word or short phrase that isn’t anywhere online. That means it shouldn’t be a pet’s name, your street, or anything you’ve posted on Facebook. Then agree on one rule: if anyone ever calls or video chats asking for emergency money, they get asked for the word.
It costs nothing. And because the word was never written down anywhere online, there’s nothing for a scammer’s AI to find.
Know someone who should see this? Share this article with your family, especially the ones who’d be first to send help if they thought you were in trouble. The code word only works if everyone knows it.