Nerd Free AI beaver mascot

Nerd Free AI

AI news and how-tos, minus the jargon

Category: AI News

  • Chrome Can Now Fill In Forms Using Your Gmail. Here’s What to Know

    Chrome Can Now Fill In Forms Using Your Gmail. Here’s What to Know

    You know that moment when a website asks for your order number, and you have to leave the page, dig through your email, copy a long string of letters and numbers, and come back? Google wants Chrome to do that digging for you.

    On October 28, Google announced that Chrome’s enhanced autofill can now pull details straight out of your Gmail. It uses Google’s Gemini AI to spot things like order numbers, package tracking info and flight bookings, then offers to fill them into forms for you. It’s handy. It also means an AI is reading your email to do it, so it’s worth knowing what’s going on and how to switch it off if you’d rather not.

    What’s new

    Chrome already fills in your name, address and saved passwords. The new part is that it can now suggest information it finds in your Gmail inbox. Google gives three examples: order numbers (say, for a return), shipping and tracking details, and flight bookings.

    When Chrome finds a match, the suggestion shows up in the usual autofill spot, labeled “Suggested by Gemini,” according to 9to5Google’s reporting. You tap it, and the box fills in. Google says the sensitive data is encrypted and that “you remain in control.”

    Google tested a version of this earlier in the year that could also pull from Google Photos and other apps. The version rolling out now sticks to Gmail.

    Who gets it

    Don’t panic if you don’t see it yet. It’s rolling out in the US on Chrome for computers, Android phones and iPhones, and only for people who use Gemini in Chrome with Google’s Personal Intelligence feature turned on. (Personal Intelligence is Google’s name for letting Gemini look at your own Google stuff, like your email, to give more personal answers.)

    Based on Google’s Gemini help pages, as reported by Notebookcheck, you also need:

    • A personal Google account. Work, school and supervised (kids’) accounts are left out.
    • To be 18 or older.
    • Gemini’s activity setting (called Keep Activity) turned on.
    • To live outside Europe’s EEA, Switzerland and the UK, where it isn’t offered.

    What this means for you

    • It can save real time. Returns, delivery problems and airline check-ins get a little less annoying when you don’t have to hunt for a confirmation email.
    • An AI is reading your inbox to make it work. That’s the trade. If your email holds things you’d rather keep away from AI, you may want to leave it off.
    • Check the privacy fine print. Notebookcheck points to a Google help page for business admins saying the default for regular users lets data help improve Google’s AI, and that it “can also be read by human reviewers” for that purpose. Android Authority also notes Google recently said Personal Intelligence data can be used to personalize ads.
    • Always glance before you hit Submit. AI can grab the wrong order or an old flight. A two-second check beats a mix-up.

    How to check it, fix it or turn it off

    If a suggestion looks wrong, or you’re just curious where it came from, Android Police reports that on Android you can press and hold the suggestion, then tap View sources to see which email it came from, or Remove suggestion to get rid of it. Removing a suggestion doesn’t delete the email itself.

    To turn the feature off in Chrome:

    • Open Chrome and go to Settings.
    • Choose Autofill and passwords.
    • Find Enhanced autofill and switch it off.

    Want Gemini to stop looking at your Gmail altogether, not just in Chrome? Open Gemini’s settings, find Connected Apps, and disconnect Gmail. You can always connect it again later.

    A quick word on safety

    AI tools that read your email and act inside your browser are new, and security researchers are still poking at them. In August, researchers showed that a single, carefully written email could trick several AI browser helpers, Gemini in Chrome among them, into doing things the user never asked for, Notebookcheck reports. That doesn’t mean this feature is unsafe. It does mean it’s smart to keep an eye on what gets filled in, and to be extra careful on banking and payment forms.

    If you’d like a quick tune-up of your accounts anyway, our 15-minute online security checkup with free AI is a good place to start. And if you use an iPhone, we also covered ChatGPT’s new way to read iCloud Mail and Calendar. For more on Google’s Personal Intelligence in the car, see Gemini in Android Auto.

    Common questions

    What is Chrome enhanced autofill with Gmail?

    It’s a new Chrome feature that uses Google’s Gemini AI to find details in your Gmail, like order numbers, tracking info and flight bookings, and suggest them when you fill out a form online.

    Is Chrome reading my email?

    Only if you use Gemini in Chrome with Personal Intelligence turned on. In that case Gemini looks through your Gmail to find details it can suggest. Google says the data is encrypted and that you stay in control.

    How do I turn off enhanced autofill in Chrome?

    Open Chrome Settings, choose Autofill and passwords, and switch off Enhanced autofill. To stop Gemini from using Gmail at all, disconnect Gmail under Connected Apps in Gemini’s settings.

    Who can use Chrome’s Gmail autofill?

    It’s rolling out in the US on computers, Android and iPhone for adults 18 and older with personal Google accounts who use Gemini in Chrome with Personal Intelligence on. Work, school and kids’ accounts are left out.

    Does removing a suggestion delete my email?

    No. Removing a suggestion only stops Chrome from offering it. The original email stays in your Gmail.

    Curious which AI gadgets are actually worth it?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • ChatGPT Can Now Read Your iCloud Email and Calendar. Here’s What to Know Before You Connect It

    ChatGPT Can Now Read Your iCloud Email and Calendar. Here’s What to Know Before You Connect It

    The short version: On Tuesday, October 27, OpenAI added two new plugins (think of them as add-ons) to ChatGPT: iCloud Mail and iCloud Calendar. If you use Apple’s email or calendar, you can now let ChatGPT search your inbox, sum up long email threads, write and even send emails, and add or change events on your calendar. It works on every ChatGPT plan, including the free one, on the web, the desktop app and iPhone. It could save you real time. It also means handing a chatbot the keys to your inbox, so here’s what it does, how to set it up, and how to stay in control.

    What ChatGPT can do with your iCloud account

    Once you connect them, here’s what each plugin lets ChatGPT do, according to OpenAI and early reports from MacRumors and Digital Trends:

    • iCloud Mail: search and read your email, catch you up on your inbox, manage messages and drafts, write replies, and send email for you.
    • iCloud Calendar: read your schedule, create new events, and update existing ones. These are the same events you see in the Calendar app on your iPhone.
    • Contacts: both plugins can look up people in your iCloud contacts, but only to read them. There’s no separate Contacts or Reminders plugin yet.

    OpenAI’s own example on X: ask ChatGPT to “find that email, catch up on your inbox, check what’s next, or update a calendar event.” So you could type something like “When is my dentist appointment, and did they email me about it?” and get one answer instead of digging through two apps.

    The fine print worth knowing

    • It only sees iCloud email. That means addresses ending in @icloud.com, @me.com or @mac.com. If your iPhone’s Mail app also shows a Gmail or Outlook account, ChatGPT can’t see those through this plugin.
    • Mail and Calendar are separate. You connect each one on its own. Turning on Mail doesn’t turn on Calendar.
    • You need a special password. Apple makes you create an app-specific password: a one-time password just for ChatGPT, so ChatGPT never gets your real Apple password. Your Apple Account needs two-factor authentication turned on (that’s the extra code Apple sends when you sign in somewhere new).
    • It can work behind the scenes too. MacRumors notes the iCloud connection can also be used by Dots, the “proactive” AI helper OpenAI launched last month, which does tasks on its own initiative.

    How to connect it (about 5 minutes)

    If you decide to try it, here are the steps, based on Apple’s and OpenAI’s help pages:

    1. Go to account.apple.com and sign in.
    2. In Sign-In and Security, choose App-Specific Passwords, then Generate an app-specific password. Name it “ChatGPT” so you remember what it’s for. Copy the password it shows you.
    3. In ChatGPT, open Settings > Plugins and find iCloud Mail or iCloud Calendar. Tap Install plugin, then Connect.
    4. For Mail, type your iCloud email address. For Calendar, type the main email address you use to sign in to Apple (it might be different). Paste the app-specific password into the connection form.
    5. Repeat for the other one if you want both.

    Important: OpenAI says to paste the password only into the connection form, never into a chat message. And never type your regular Apple password into ChatGPT.

    What this means for you

    • It’s optional. Nothing changes unless you connect it yourself. If you’d rather keep your email away from AI, just don’t.
    • Read before it sends. ChatGPT can send email as you. OpenAI says it may ask for your approval before reading information or taking an action. Always read the draft, check who it’s going to, and only then say yes.
    • Double-check dates and times. AI can mix up details. Before you rely on an appointment it finds or creates, glance at your Calendar app.
    • Watch out for fake setup help. The only places you need are account.apple.com and ChatGPT’s own Settings. Anyone who texts, emails or calls offering to “connect ChatGPT to your iCloud” for you is likely a scammer.
    • Know how to turn it off. In ChatGPT, go to Settings > Plugins, pick the plugin and choose Disconnect. Then also go to account.apple.com, open App-Specific Passwords and remove the “ChatGPT” one. OpenAI says disconnecting in ChatGPT alone does not cancel the password at Apple.
    • Changing your Apple password resets it. Apple says changing or resetting your main Apple password cancels all app-specific passwords, so you’d need to make a new one to reconnect.

    The bigger picture

    AI helpers are racing to get into the places where our lives actually happen: email, calendars, banks, cars. Just this month, Apple gave its own Siri more access to your messages and voicemails, and Google put Gemini in the car. ChatGPT reaching into iCloud is part of the same trend. The upside is less busywork. The trade-off is that one company’s chatbot now knows a lot more about you. A good rule of thumb: connect only what you’d be comfortable with a helpful stranger reading, and start small. You can always add more later.

    Want to read more? See how Siri now remembers your chats and has daily AI limits, how Gemini in your car can check your email and calendar, or take 20 minutes for an online security checkup with free AI.

    Common questions

    Can ChatGPT read my iCloud email now?

    Yes, if you connect it. OpenAI added iCloud Mail and iCloud Calendar plugins on October 27, 2026. Once connected, ChatGPT can search and read your iCloud email, write replies and send messages, and read, create and update calendar events.

    Is the iCloud plugin free?

    Yes. OpenAI says the iCloud Mail and Calendar plugins are available on all ChatGPT plans, including the free plan, on the web, the desktop app and iPhone.

    Can ChatGPT see my Gmail through the iCloud plugin?

    No. The iCloud Mail plugin only works with Apple email addresses ending in @icloud.com, @me.com or @mac.com. Gmail or Outlook accounts added to the iPhone Mail app are not included.

    Do I have to give ChatGPT my Apple password?

    No, and you shouldn’t. Apple requires an app-specific password, a separate password you create at account.apple.com just for ChatGPT. You need two-factor authentication turned on to make one.

    How do I disconnect ChatGPT from iCloud?

    In ChatGPT, open Settings, then Plugins, choose the iCloud plugin and select Disconnect. Then sign in at account.apple.com, open App-Specific Passwords and remove the one you made for ChatGPT, because disconnecting in ChatGPT does not cancel it at Apple.

    Curious what else AI can help with?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • iPhone Update: Siri Now Remembers Your Chats, but Apple Is Adding Daily AI Limits

    iPhone Update: Siri Now Remembers Your Chats, but Apple Is Adding Daily AI Limits

    The short version: Apple released iOS 27.2 on Monday, October 26. If you have a newer iPhone, the update makes the new Siri (Apple calls it Siri AI) a lot more like a helpful assistant: it can now remember your past conversations and pull answers from more of your stuff, like your voicemails and photos people send you in Messages. But tucked into the fine print is a catch: Apple now says Siri AI and some other AI features have daily usage limits, and that more access will cost money someday. Here’s what changed, in plain English, and the settings worth checking.

    What’s new for Siri AI in iOS 27.2

    Siri AI is Apple’s rebuilt, smarter Siri. It’s still labeled a beta (a test version), and it only works on an iPhone 15 Pro, iPhone 15 Pro Max, or any iPhone 16 or newer. If your phone is older, you’ll still get iOS 27.2’s other changes, just not these Siri ones. According to Apple’s release notes, the update adds:

    • Memory of past chats. You can ask Siri about something it told you before. Apple’s own example: “the restaurants it recommended last week.”
    • More places to look for answers. Siri can now use photos attached in Messages, the people invited to your Calendar events, your voicemails, and your Safari history and Reading List.
    • One switch to keep apps private. A setting called Excluded Apps lets you pick which apps Siri, Search and suggestions can use, all in one place.
    • Copy just part of an answer. You can now select and copy only the piece of a Siri reply you need, instead of the whole thing.
    • More languages. Siri AI now works in French, Japanese, Korean, Portuguese and Spanish, as well as English. (It’s still not available in the EU or mainland China, and not for kids under 13.)

    Other handy bits in the same update: Messages can suggest actions based on your chats (like adding a dinner plan to your calendar), the Apple TV app gets separate profiles for each person in the house, and the Home app can automatically lock compatible smart locks after the door closes.

    The catch: daily limits, and a fee “in the future”

    In the small print of its announcement, Apple says some Apple Intelligence features that run on Apple’s computers (instead of on your phone) now have daily usage limits. That includes Siri AI, the smart photo editing tools, Image Playground (Apple’s picture maker) and some Shortcuts features. Apple says the limits can change depending on the feature, how complicated your request is, and how busy its systems are.

    Then comes the sentence people noticed: “Expanded access to such features will be available for a fee in the future.” In other words, the basics stay free, but heavy users may eventually be asked to pay for more. Apple hasn’t said how much it will cost, when it starts, or how much you get free each day. Apple’s usage terms say that if you hit a limit, that one feature takes a break until the limit resets, and everything else on your phone keeps working.

    This isn’t totally new. Back in June, Apple said iCloud+ subscribers would get higher limits for some features, like making images. What’s new is seeing Siri itself on the limited list. It’s a sign of where AI helpers are heading: a free amount for everyday use, and a paid tier for people who use them a lot. (ChatGPT and Google’s Gemini already work this way.)

    What this means for you

    • Should you update? Yes. iOS 27.2 also includes security fixes, and those matter more than any new feature. Go to Settings > General > Software Update. Plug in and use Wi-Fi first.
    • Don’t worry about the limits yet. Asking Siri for the weather, a timer or a quick question a few times a day is very unlikely to hit a cap. If a feature stops working for the day, just try again later.
    • Keep private apps away from Siri. Go to Settings > Siri > Excluded Apps, tap Add Apps, and pick apps like your banking, health or journal apps. They’ll stay out of Siri, Search and suggestions. (It doesn’t delete the app or log you out.)
    • Decide how long Siri remembers. Go to Settings > Siri > Keep Conversations and choose 30 days or 1 year. Older chats are then permanently deleted.
    • Delete a chat you don’t want saved. Open the Siri app, touch and hold a conversation, tap Delete, then confirm.
    • Don’t pay anyone for “Siri Plus” yet. Apple hasn’t announced a price or plan. Any ad, text or email selling one right now is a scam. Real Apple charges only show up in Settings on your iPhone.

    The bigger picture

    AI assistants are expensive to run. Every time you ask a smart question, a big computer somewhere does a lot of work. Companies are figuring out who pays for that, and the answer is increasingly “you, if you use it a lot.” For most people, the free amount will probably be plenty. The bigger question is privacy: an assistant that remembers your chats and reads your voicemails is more useful, but it also knows more about you. The good news is that Apple gives you the controls. It’s worth five minutes to set them the way you like.

    Curious how other AI helpers handle this? See our story on free Gemini users losing the model picker, or learn how to do a 20-minute online security checkup with free AI.

    Common questions

    What is new in iOS 27.2 for Siri?

    On an iPhone 15 Pro or newer, Siri AI can now remember past conversations, use more of your information (like voicemails, photos in Messages and Calendar attendees), and copy part of an answer. A new Excluded Apps setting lets you choose which apps Siri can use. Siri AI also now works in French, Japanese, Korean, Portuguese and Spanish.

    Does Siri AI have a daily limit now?

    Yes. Apple says Siri AI, its smart photo editing tools, Image Playground and some Shortcuts features have daily usage limits that can vary. If you reach one, that feature pauses until the limit resets, while the rest of your phone keeps working.

    Will I have to pay for Siri?

    Not right now. Apple says expanded access to these AI features will be available for a fee in the future, but it has not announced a price, a start date or how much stays free. Be wary of anyone selling a paid Siri plan today.

    How do I stop Siri from using a certain app?

    Open Settings, tap Siri, then Excluded Apps, and tap Add Apps. Choose the apps you want to keep out of Siri, Search and suggestions.

    How do I delete my Siri conversations?

    Open the Siri app, touch and hold a conversation, tap Delete and confirm. To delete old chats automatically, go to Settings, then Siri, then Keep Conversations, and choose 30 days or 1 year.

    Curious what else AI can do around the house?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • Is Your Samsung TV Listening? What a Researcher Claims, and How to Turn Off TV Tracking

    Is Your Samsung TV Listening? What a Researcher Claims, and How to Turn Off TV Tracking

    The short version: A security researcher says they cracked open their own 2025 Samsung TV and found it using on-device AI to collect a surprising amount of information for ads: what you watch, which ads you skip, screenshots of the screen and even labels for sounds in the room. The claims have not been independently confirmed yet, and Samsung hasn’t publicly responded in the reports we read. But the tracking settings it points to are real, and you can turn most of them off in about five minutes. Here’s what was claimed, what we actually know, and what to tap.

    What the researcher says they found

    The researcher goes by log4j on X. On October 21 they posted that they kept getting “weirdly targeted ads” on their 2025 Samsung TV. The TV’s software was locked with encryption, so they found their own security holes to get full control of it (hackers call this “rooting”). Then they live-posted what they say they found inside Tizen, the software Samsung TVs run on.

    According to their posts, as reported by Tom’s Hardware, Notebookcheck and HDTVTest, the TV was:

    • Tracking viewing habits, such as which videos you watch, how far you get and which ads you skip.
    • Sorting sounds into categories, like a dog barking or a baby crying, rather than saving full recordings of what people say.
    • Taking screenshots and running them through computer vision, the kind of AI that recognizes what’s in a picture.
    • Tying it to an advertising ID that can follow you across devices, so the ads you see elsewhere can match what you watched.
    • Borrowing a microphone from a connected soundbar or Bluetooth gadget if the TV itself has none. This one is the claim that got the most attention, and it’s also one of the hardest to check.

    The researcher’s summary was a joke: they “finally found a company that truly listens to its customers.” They say a full write-up is coming, but it hadn’t been published when we wrote this.

    What we actually know so far

    This is the part to read slowly. Everything above comes from one person’s posts. No other lab has repeated the full findings yet. A Hugging Face engineer named ngxson tried to test the microphone part on their own rooted Samsung TV. They saw the mic switch on for exactly five seconds, then stay off for hours. They also pointed out that their TV already had several settings turned off, so a regular TV fresh out of the box might act differently.

    Some of these features also have a friendlier side. Samsung advertises sound-sensing features on its 2025 Vision AI TVs, called Pet Care and Family Care, that can alert your phone through the SmartThings app if, say, your dog keeps barking while you’re out. Hearing “dog barking” detected inside the TV doesn’t prove it’s being used for ads. What’s missing is a clear answer from Samsung about what gets collected, when, and where it goes.

    One thing isn’t new at all: most smart TVs, from Samsung and other brands, use something called ACR (automatic content recognition). It quietly identifies what’s on your screen, even from cable or a game console, and uses that to target ads. Other TV makers, including LG, have faced similar spying complaints. That’s exactly the setting you can switch off below.

    What this means for you

    • Don’t panic, but do take five minutes. The scariest claims are unproven. The basic ad tracking is real either way, and it’s easy to turn down.
    • Turn off “Viewing Information Services.” On most 2022 and newer Samsung TVs, go to Settings > All Settings > General & Privacy > Terms & Privacy and untick Viewing Information Services. That’s Samsung’s ACR. Menu names shift a little by model year, so look for “Privacy” if you don’t see it.
    • Turn off “Interest-Based Advertising” on the same screen. You’ll still see the same number of ads; they just won’t be built from your viewing.
    • Check your Samsung account too. If you sign in to a Samsung account, look in its privacy settings for an option about personalized ads using your account data and switch it off.
    • Recheck after updates. Privacy groups say big software updates can flip these back on. A quick look every few months is smart.
    • Think about what’s paired. If the soundbar claim worries you, you can unpair Bluetooth gadgets you don’t use, and turn off any hands-free “wake up” voice option in your TV’s voice settings.
    • Use it offline if you only watch through a streaming stick or cable box. A TV that isn’t connected to Wi-Fi can’t send data home. (The stick or box may have its own tracking settings, so check those as well.)

    The bigger picture

    TVs are cheap partly because ads and data pay part of the bill. AI makes that data far more detailed, because a TV that can “see” and “hear” can learn a lot about a home. We’ll update you if the researcher’s full report comes out or Samsung responds. In the meantime, the same “check your settings” habit helps everywhere: our 20-minute online security checkup covers your phone and accounts, and we explained how ChatGPT ads use your email and phone number for targeting.

    Common questions

    Is my Samsung TV listening to me?

    Not proven. A security researcher claims 2025 Samsung TVs use on-device AI to label sounds in the room, like a dog barking, and to take screenshots for ad targeting. Those claims haven’t been independently confirmed, and one engineer who tested a modified TV saw the microphone turn on for only five seconds.

    How do I stop my Samsung TV from tracking what I watch?

    On most 2022 and newer Samsung TVs, go to Settings, All Settings, General and Privacy, then Terms and Privacy. Untick Viewing Information Services and Interest-Based Advertising. Menu names vary a little by model, and big software updates can reset them, so recheck now and then.

    What is Viewing Information Services on a Samsung TV?

    It’s Samsung’s name for automatic content recognition, or ACR. It identifies what is showing on your screen, including from cable boxes and game consoles, and that information can be used to recommend shows and target ads. You can turn it off in the TV’s privacy settings.

    Will turning off ad tracking remove ads from my TV?

    No. You’ll usually see the same number of ads on the home screen and in free channels. They just won’t be personalized based on what you watch.

    Is it only Samsung TVs?

    No. Most smart TV brands use some kind of viewing tracking for ads, and LG has faced similar complaints. The new AI claims are specific to Samsung’s 2025 TVs, but every smart TV owner should check their privacy settings.

    Want gadgets that work for you, not the ad business?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • Big Banks Are Blocking AI Assistants From Your Account. Here’s What That Means for You

    Big Banks Are Blocking AI Assistants From Your Account. Here’s What That Means for You

    The short version: AI “agents” are the new kind of assistant that doesn’t just answer questions but actually does things for you, like booking a table or paying a bill. Some people have started handing these agents their bank username and password. This week, America’s biggest banks, including JPMorgan Chase, Bank of America, Wells Fargo, Citi and Morgan Stanley, made it clear they’re not OK with that, and they’re blocking or limiting AI agents that try to log in to customer accounts. Here’s what’s going on, and what it means for your money.

    First, what’s an AI agent again?

    A regular chatbot is like a very smart pen pal: you ask, it answers. An AI agent is more like a personal assistant you hand your keys to. You give it a goal (“pay my electric bill” or “find me a cheaper flight and book it”), and it opens websites, clicks buttons and fills in forms on its own. They’re catching on fast: Meta’s agent, Muse, hit 10 million users on October 21.

    That’s handy for restaurant reservations. It’s a lot scarier when the “keys” are the ones to your checking account.

    What the banks found

    On Thursday, October 22, the news site Axios reported that major U.S. banks are spotting unvetted AI agents operating inside customer accounts. In most cases, a customer had given the agent their own login and password. But some agents also called bank customer service lines, and some appeared to take actions that went beyond what the customer asked for, according to the report.

    Part of the problem is that banks can’t easily tell who’s who. Is that login really you? Is it an agent you approved? Or an agent that has wandered off script? Andre Durand, the CEO of the security company Ping Identity, told USA Today this is a “window of vulnerability.” He added that the old trick of emailing you a one-time code isn’t enough anymore, because many AI agents can read your email too.

    What each bank is doing

    • JPMorgan Chase blocks outside AI agents from working on behalf of everyday customers. “You don’t want rogue agents,” CEO Jamie Dimon said on October 13. A Chase spokesperson’s advice: don’t give your username and password to anyone.
    • Bank of America is restricting AI agents’ access to customer deposits. CEO Brian Moynihan said he wants “a human being in the loop,” meaning a real person approves a payment before it happens.
    • Citi told USA Today “We do not permit third-party AI agents or assistants to log into a customer’s account or transact directly.”
    • Wells Fargo says it has blocked unauthorized agents.
    • Morgan Stanley and the brokerage Charles Schwab are blocking outside agents too. Schwab has its own in-house assistant, called Charley, that can answer questions but can’t make trades.

    Most banks are open to a middle ground: read-only access. That means an approved app can look at your balance and transactions, the way budgeting apps such as Quicken already do through secure connection services like Plaid, but it can’t move your money.

    The fine print to watch: who pays if an agent messes up?

    Here’s the part that could affect your wallet. Wells Fargo has updated its online banking agreement, starting November 24, 2026, with a new section on AI agents. According to USA Today, it says that if you let an AI agent into your account, you’re “solely responsible” for supervising it, and transactions it makes, “including those you did not want or intend,” count as authorized by you.

    In plain English: if you hand an AI assistant your bank login and it pays the wrong person, the bank may treat it like you did it. Getting your money back could be much harder than if a stranger had stolen your card. Other banks may add similar rules, so it’s worth reading those “we’ve updated our terms” emails this season instead of tossing them.

    What this means for you

    • Don’t give any AI assistant your bank username and password. Not ChatGPT, not Muse, not a shiny new app. Your bank says so, and you could be on the hook for anything it does.
    • Let AI look, not touch. It’s fine to ask a chatbot to explain a fee or help you plan a budget. Just type in the numbers yourself, without account numbers or logins. Our guide to making a monthly budget with free AI shows how.
    • Use your bank’s own tools first. Many banks have built-in assistants, bill pay and alerts that don’t require sharing your password with anyone.
    • Turn on account alerts. A text or app alert for every withdrawal or payment over a small amount helps you catch anything odd fast, whether it’s an AI mix-up or a plain old crook.
    • Use a passkey or two-step login on your bank account if it’s offered. Security experts say passkeys and fingerprint or face logins help prove that a real human is the one signing in. Our 20-minute online security checkup walks you through it.
    • Watch out for scams riding this headline. A real bank won’t call or text asking you to “verify your AI assistant” or read back a code. Hang up and call the number on the back of your card.

    The bigger picture

    This isn’t the end of AI helpers in banking. Banks know customers will want them, and several are building their own or working on safer ways to let approved agents in, with you approving each payment. For now, though, the message from the people who hold your money is simple: AI can be your helper, but it shouldn’t have your keys. It’s the same tug-of-war we covered when big websites started blocking AI agents and when Walmart opened its doors to AI shopping agents.

    Common questions

    Are banks blocking AI agents from customer accounts?

    Yes. In October 2026, JPMorgan Chase, Bank of America, Wells Fargo, Citi, Morgan Stanley and Charles Schwab said they block or limit third-party AI agents that log in to customer accounts or try to move money. Many still allow approved apps to have read-only access to things like your balance.

    Is it safe to give an AI assistant my bank password?

    No. Banks such as Chase say you should never share your username and password with anyone, including AI assistants. If an agent you gave access to makes a payment you didn’t want, some banks may treat it as authorized by you, which can make getting your money back harder.

    What is Wells Fargo’s new rule about AI agents?

    Wells Fargo’s updated online access agreement, which takes effect November 24, 2026, says customers who let an AI agent into their account are solely responsible for supervising it, and that transactions it makes, including ones you did not want or intend, are considered authorized.

    Can I still use AI to help with my money?

    Yes, as long as the AI doesn’t have your login. You can ask a free chatbot to explain fees, build a budget or compare options using numbers you type in yourself. For anything that moves money, use your bank’s own app or website.

    What is read-only access?

    Read-only access means an approved app can see information like your balance and transactions but can’t send money, pay bills or make changes. Banks already allow this for budgeting apps through secure connection services, and many are moving toward allowing AI tools the same limited access.

    Want tech that makes everyday life easier?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • AI Helped Find a Huge Hole in the Code Behind Most Websites. Here’s What It Means for You

    AI Helped Find a Huge Hole in the Code Behind Most Websites. Here’s What It Means for You

    The short version: On Wednesday, October 21, 2026, the people who maintain PHP, the behind-the-scenes code that runs roughly 7 out of 10 websites, rushed out fixes for two very serious security holes. The bugs, nicknamed PHP2Shell, were found with help from AI. Within hours, websites around the world, from DeviantArt to the Internet Archive to government sites, went offline for a while to patch up. The good news: most regular people don’t need to do anything. The people who need to act are website owners. Here’s what happened, in plain English.

    First, a quick vocabulary check

    • PHP is a programming language, a set of instructions computers understand. It quietly powers a huge share of the web, including WordPress, the tool behind millions of blogs and small business sites. You never see it. It works in the back room, like the kitchen of a restaurant.
    • A server is the computer, usually sitting in a big warehouse somewhere, that holds a website and sends it to your screen.
    • A vulnerability (or “security hole”) is a mistake in software that a bad guy could use to break in.
    • A patch is the fix. Think of it as the repair crew showing up with a new lock.
    • “Shell” is techie slang for full control of a computer. “PHP2Shell” means a hole in PHP that could let an attacker take over the server. That’s why everyone moved so fast.

    What actually happened

    Two flaws, with the official tracking numbers CVE-2026-109618 and CVE-2026-109619, were rated 9.2 out of 10 for severity. According to security company Tenable, every version of PHP going back to version 5.4 from 2012 is affected by at least one of them. In the worst case, an attacker could send a specially made request to a website and run their own commands on its server, without needing a password.

    The fixes came out on October 21 as PHP versions 8.5.12, 8.4.27, 8.3.36 and 8.2.35. The release was even moved up a day after the bug numbers leaked early. Older versions (PHP 8.1 and below) are past their official support dates, so they didn’t get an official update. Owners of those sites are basically being told: upgrade, or protect it some other way.

    Where AI comes in

    This is the part that makes it AI news. Two separate teams found the bugs, and both had AI helpers in the passenger seat:

    • A researcher at a security firm called Calif found them using an OpenAI model through an OpenAI program called “Patch the Planet,” which puts AI to work hunting for bugs in widely used software.
    • Cloud security company Wiz found both flaws on its own, using its AI security system along with AI models from Google and OpenAI. Wiz said building the full attack was “at the edge of what human researchers could practically achieve,” and called it “among the most severe vulnerabilities discovered” in software history.

    OpenAI called PHP2Shell “one of the most impactful security issues in recent memory.” It also warned that it expects the number of AI-found bugs to “increase sharply in the coming months.” Translation: AI is getting very good at spotting weak spots in code. That’s great when the good guys find them first, like they did here, and a little nerve-racking when you think about who else has these tools.

    Why websites went dark

    Government security agencies didn’t mince words. The U.S. Cybersecurity and Infrastructure Security Agency (CISA) told organizations to patch immediately, or take affected systems offline until they could. Similar warnings came from the NSA and the UK’s National Cyber Security Centre.

    So a lot of websites pulled the plug for a bit. Over the next day or so, outages or maintenance pages showed up at Wikimedia, DeviantArt, the Internet Archive, Pokémon Home, some BBC pages, and government websites in Finland, the Netherlands, and Scotland, according to the BBC and The Verge. A couple of older online games in Japan even announced they’d shut down for good rather than fix their aging code.

    Meanwhile, the bad guys got busy. Cloudflare, a company that sits in front of millions of websites to protect them, said attack attempts jumped from thousands to millions within hours of the news. Cloudflare turned on extra protection for all its customers, including free ones, to buy time. As of the latest reports, the PHP Foundation (the group that looks after PHP) said it didn’t know of any successful break-ins using these bugs, but security firms say attackers are actively trying.

    What this means for you

    • If you just use websites, you probably don’t need to do anything. This is a bug in the websites’ computers, not in your phone or laptop. There’s no app to update on your end.
    • If a site you like is down, give it time. A “down for maintenance” page this week likely means the owners are doing the responsible thing.
    • Watch for “please change your password” emails from real companies. If a site you use tells you it was affected, change that password, and change it anywhere else you used the same one. Go to the site yourself instead of clicking the email link.
    • Be extra careful with scary emails about this news. Scammers love a headline. Ignore messages that say “Your account was hacked by PHP2Shell, click here.” Not sure about a message? Here’s how to ask AI if a text or email is a scam.
    • If you own a website (a blog, a shop, a church or club site), ask your hosting company or web person one question: “Have you updated PHP for the PHP2Shell fixes from October 21?” Updating WordPress or your plugins does not fix this one. It has to be fixed on the server, which is usually your host’s job.
    • It’s a great week for a security tune-up. Our new 20-minute online security checkup with free AI walks you through the basics, step by step.

    Try this next

    Want to stay a step ahead of the bad guys? Learn how Android’s AI can flag scam texts and calls, or read why some websites are blocking AI agents.

    Common questions

    What is PHP2Shell?

    PHP2Shell is the nickname for two critical security flaws (CVE-2026-109618 and CVE-2026-109619) in PHP, the programming language that runs a large share of websites, including WordPress sites. In the worst case, they could let an attacker take control of a website’s server. Fixes were released on October 21, 2026.

    Do I need to do anything about PHP2Shell?

    Most people who just visit websites don’t need to do anything, because the flaw is on websites’ servers, not your devices. Do change your password if a company tells you it was affected, and watch out for scam emails that use the news. Website owners should ask their host if PHP has been updated.

    Will updating WordPress fix PHP2Shell?

    No. PHP2Shell is a flaw in PHP itself, which runs underneath WordPress on the server. Updating WordPress or plugins won’t fix it. Your hosting company or web developer needs to update PHP to a fixed version (8.5.12, 8.4.27, 8.3.36 or 8.2.35) or newer.

    How did AI help find the PHP flaws?

    Two teams found the bugs separately. A researcher at the security firm Calif used an OpenAI model through OpenAI’s Patch the Planet program, and the cloud security company Wiz used its own AI system along with Google and OpenAI models. Both reported the bugs privately so they could be fixed before details went public.

    Why were so many websites down this week?

    U.S. and UK security agencies told organizations to install the PHP fixes right away or take affected systems offline. Many sites, including DeviantArt, the Internet Archive and some government sites, went offline for a while to patch, while attackers began scanning the web for sites that hadn’t been fixed yet.

    Want tech that makes everyday life easier?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • Walmart Will Let AI Assistants Shop for You, with Spending Limits You Set. Here’s How It Works.

    Walmart Will Let AI Assistants Shop for You, with Spending Limits You Set. Here’s How It Works.

    The short version: On Wednesday, October 21, 2026, Walmart announced new tools that let AI assistants shop at Walmart and Sam’s Club for you, with rules to keep you in charge. The big one for regular shoppers is called Agent Profiles, coming within the next month. It gives each AI helper you use its own spot in your Walmart account, so you can see exactly what it bought, require your OK before every purchase, or set a spending limit like $50 or $100. Walmart also says its normal returns process applies even if the AI messes up. Here’s what’s new, why it matters, and what to think about before you let a robot fill your cart.

    First, what is an “AI agent”?

    An AI agent is an AI helper that doesn’t just answer questions. It actually does things for you, like searching a store, putting items in a cart, and checking out. You might tell one, “Reorder my usual paper towels and dog food,” and it goes off and handles it.

    Walmart already works with several of these. The company says it connects with AI apps including ChatGPT, Google’s Gemini, and Meta’s Muse, plus its own shopping assistant, Sparky. The new tools are about making that kind of shopping safer and easier to keep an eye on.

    The problem Walmart says it’s fixing

    Right now, when an AI agent shops for you, it signs in as you. According to Walmart, that means everything lands in one big pile. Your purchases and the AI’s purchases look exactly the same, so if something odd shows up, it’s hard to tell who ordered it. (Was it you at midnight, or the robot? Honestly, both are plausible.)

    What’s new: three pieces

    1. Agent Profiles (for shoppers)

    Walmart says Agent Profiles will launch “in the next month.” Here’s how they work, based on Walmart’s announcement:

    • Each AI helper gets its own profile inside your Walmart account. Its orders show up separately from yours.
    • You choose how purchases get approved. You can require your OK for every single purchase, or let the agent buy things on its own up to a spending limit you pick. Walmart gives $50, $100, or $200 as examples.
    • You can delete a profile anytime, which cuts that agent off.
    • No Walmart+ membership needed. TechCrunch reports agent shopping works for anyone with a Walmart account.

    Walmart compares it to adding someone as an authorized user on your credit card: they can buy things, you set the limit, and you can see what they spent.

    Getting started is meant to be simple. Walmart says the first time you use an AI agent to shop through its new system, a profile gets created for that agent automatically. You then sign in to your Walmart account to link it and set your limits.

    2. The Walmart Agent Gateway (behind the scenes)

    Today, AI agents shop by clicking around Walmart’s website like a person would, which Walmart says can be slow and error-prone. The Agent Gateway is a direct, behind-the-scenes connection built just for AI agents. It uses a standard called MCP, a common way for AI tools to plug into apps and stores. It’s starting as a test for companies and developers, so you won’t see it directly.

    One nice touch: Walmart says the gateway can help an agent find your “usuals,” like the brand of milk you always buy, without handing the agent your whole shopping history.

    3. Rules that AI agents must follow

    Walmart also published terms that agents have to play by. Among them:

    • You must approve agent purchases by default, and Walmart may ask for extra approval even when a purchase is under your limit.
    • Agents can’t do more than what you intended.
    • Agents never get your card number, and they can’t add or change your payment methods.
    • Agents only get the information needed for the task, and only what you allow.
    • Walmart’s returns process applies, including when an agent doesn’t do what you wanted.
    • Agents that break the rules get blocked from Walmart’s site.

    Walmart also says AI agents see the same prices a person would see on Walmart.com. In the company’s words, “We price products, not people.”

    Why this is a big deal right now

    Stores are split on AI shoppers. As we covered recently, some big websites are blocking AI agents. TechCrunch reports that Amazon, Walmart’s biggest rival, recently started blocking AI agents from making purchases for shoppers, and some travel and booking sites have locked agents out completely. Walmart is going the other way and rolling out the welcome mat, as long as the agents follow its rules.

    For you, that means where you shop may decide whether your AI assistant can help. If you’re curious about AI shopping help, Walmart is about to become one of the easier places to try it, with guardrails built in.

    What this means for you

    • You don’t have to do anything. Walmart says shopping with an agent “won’t be for everyone.” Your regular Walmart account works exactly the same if you never touch this.
    • If you try it, start with “approve every purchase.” Let the AI build the cart, and you press the final button. Once you trust it, you can switch to a small spending limit.
    • Keep the limit small. A $50 limit is a lot less scary than a $200 one if the AI misreads “two” as “twenty.”
    • Check the agent’s order list now and then. That’s the whole point of Agent Profiles: you can see exactly what it bought.
    • Know your exit. You can delete an Agent Profile anytime, and Walmart says returns still apply if the AI gets it wrong.
    • Remember the AI company sees things too. Walmart notes the AI app’s own privacy policy covers what you share with that app. Only connect agents from companies you trust.
    • Watch for fakes. Scammers love new features. Walmart says you link an agent by signing in to your Walmart account. Be wary of texts or emails with links asking you to “set up your Walmart AI shopper.” Go to the Walmart app or website yourself instead.

    The bottom line

    AI shopping helpers are coming whether we’re ready or not, and Walmart is betting that people will be more comfortable with them if there are clear limits, a separate receipt pile, and an easy undo button. That’s a sensible approach. Just remember the golden rule of any new shopping tech: start small, watch closely, and never let anything spend more than you’d be okay losing in a mix-up. Want to compare? Read how Google’s Universal Cart watches prices across stores.

    Common questions

    What are Walmart Agent Profiles?

    Agent Profiles give each AI shopping assistant you use its own profile inside your Walmart account. You can see its purchases separately from yours, require approval for every purchase or set a spending limit, and delete the profile anytime. Walmart says they launch within a month of its October 21, 2026 announcement.

    Can an AI agent see my credit card number at Walmart?

    No. Under Walmart’s new rules for AI agents, agents don’t receive your payment card information and can’t add or change your payment methods.

    What if an AI agent buys the wrong thing at Walmart?

    Walmart says its normal returns process applies, including when an agent doesn’t act as intended. By default, you also have to approve agent purchases before they go through.

    Do I need Walmart+ to use an AI shopping agent?

    No. TechCrunch reports that agent shopping support is available to anyone with a Walmart account, and a Walmart+ membership isn’t required.

    Curious about AI that helps in everyday life?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • Gemini in Your Car Can Now Check Your Email, Calendar, and Photos for You

    Gemini in Your Car Can Now Check Your Email, Calendar, and Photos for You

    The short version: On Tuesday, October 20, 2026, Google started rolling out a feature called Personal Intelligence for Gemini, its AI assistant, on Android Auto. If you turn it on, you can ask things like “What’s my flight gate, and is my flight on time?” or “Take me to my doctor’s appointment,” and Gemini will dig through your own Gmail, Google Calendar, Google Photos, and Google Maps to find the answer. No scrolling through emails at a red light. It’s optional, you pick which apps it can see, and you can switch it off anytime. Here’s how it works, who gets it, and what to think about before you say yes.

    First, a quick vocabulary check

    • Gemini is Google’s AI assistant, the chatbot that’s replacing the older Google Assistant on Android phones. You talk to it like a person, and it talks back.
    • Android Auto puts your Android phone’s apps, like maps, music, and messages, on your car’s dashboard screen when you plug in or connect wirelessly. (It’s Google’s version of Apple CarPlay.)
    • Cars with Google built-in are newer models that have Google’s software installed right in the car, so they work even without a phone plugged in.
    • Personal Intelligence is Google’s name for letting Gemini look at your own information in other Google apps so its answers fit you, not just anybody.

    Google says Gemini itself is now available in cars worldwide on Android Auto and on cars with Google built-in. The new part is the personal touch.

    What you can ask it to do

    Google gave a handful of examples of the kind of thing Gemini can now handle with your permission:

    • Two questions at once. On the way to the airport: “What’s my flight gate? And is my flight on time?” Gemini finds the confirmation email and checks for delays in one go.
    • Drive to an appointment without the address. “Take me to my doctor’s appointment.” Gemini pulls the address from your calendar.
    • Find a place from a photo. “Find the car wash from the photo I took last week and navigate there.”
    • Pick up where you left off. “Take me to the coffee shop we were talking about yesterday.”
    • Suggestions that fit your habits. “Find me an indoor restaurant along my route that matches my usual vibe.” Google says if you’re vegetarian, the shortlist will reflect that, so you don’t have to say it every time.

    The apps Gemini can draw from are Google Workspace (which includes Gmail and Google Calendar), Google Photos, Google Maps, YouTube, and Search. In other words, it can now answer the question “Where am I supposed to be?” better than most of us can at 7 a.m.

    Who gets it, and when

    • Android Auto: rolling out starting October 20, 2026. Like most Google features, it arrives gradually, so it may take a while to show up for you.
    • Cars with Google built-in: “in the coming months,” according to Google. In those cars, personal answers only work when you’re in the car with your phone connected.
    • Who’s eligible: personal Google accounts for people 18 and older. Work and school accounts (Google Workspace business, enterprise, and education accounts) aren’t included yet.
    • Where: worldwide, except the European Economic Area, Switzerland, the United Kingdom, and Nigeria.

    It’s opt-in, with one catch

    Google says your apps aren’t connected until you choose to turn Personal Intelligence on. Here’s the catch that tech site 9to5Google pointed out: if you already turned it on in the Gemini app on your phone, it comes with you into the car automatically. So if you said yes to it months ago and forgot, your car may soon know about your dentist appointment, too.

    If you don’t use Personal Intelligence, Google says everything else still works. You can navigate, play music, make calls, and ask Gemini general questions. The answers just won’t be tailored to you.

    How to check or change your setting

    You control this from the Gemini app on your phone, not from the car screen. According to Google’s help pages:

    1. Open the Gemini app on your phone.
    2. Tap Menu, then your profile picture or initial.
    3. Tap Personal Intelligence, then Connected Apps.
    4. Turn individual apps (like Google Workspace, Google Photos, or YouTube) on or off.

    Two fine-print details worth knowing: Gemini can still use public information from Google Search and YouTube even if you disconnect those apps. And disconnecting an app doesn’t delete your past Gemini chats. Google’s help page says you can manage or delete your Gemini Apps activity separately, and turn off the setting that keeps your future chats for improving Google’s AI.

    What this means for you

    • It can genuinely keep your hands on the wheel. Asking out loud beats thumbing through emails at a stoplight. Anything that keeps your phone in its holder is a win.
    • Do your setup in the driveway. Turn it on (or off) and pick your apps while you’re parked, not on the highway.
    • Check whether it’s already on. If you’ve used Personal Intelligence in the Gemini app before, it may already carry over to your car. Take 30 seconds to look at your Connected Apps list.
    • Think about who else rides with you. Gemini answers out loud. If you share the car, or carpool with coworkers, a question about your calendar might get read to the whole car.
    • Double-check the important stuff. AI assistants can grab the wrong email or an old appointment. Before a flight, glance at the airline’s app or the airport screens once you’re parked.
    • Connect only what you need. If you just want calendar and maps help, you can leave Photos and YouTube off.

    Try this next

    Curious what other AI helpers are learning to do for you? Read about Yelp’s AI that calls restaurants to check the wait, and how Gemini picks how hard to think for free users. And if a warning light pops up on your dashboard during that drive, here’s how to figure out what it means with free AI.

    Common questions

    What is Gemini Personal Intelligence on Android Auto?

    It’s an optional feature, rolling out since October 20, 2026, that lets Google’s Gemini assistant use your own Gmail, Google Calendar, Google Photos, Google Maps, YouTube, and Search info to answer questions while you drive, like finding your flight gate or the address of your next appointment.

    Is Personal Intelligence turned on automatically in my car?

    No, it’s opt-in. But Google says if you already turned it on in the Gemini app on your phone, that choice carries over to Android Auto automatically. You can check in the Gemini app under Personal Intelligence, then Connected Apps.

    How do I turn off Personal Intelligence for Gemini?

    In the Gemini phone app, tap Menu, then your profile picture or initial, then Personal Intelligence, then Connected Apps, and switch off the apps you don’t want Gemini to use. Everything else in the car keeps working without it.

    Who can use Gemini Personal Intelligence in the car?

    It’s rolling out to personal Google accounts for adults 18 and older, starting on Android Auto, with cars that have Google built-in coming in the next few months. It’s not available yet for work or school accounts, or in the European Economic Area, Switzerland, the UK, or Nigeria.

    Want tech that makes everyday life easier?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • Yelp’s New AI Will Call a Restaurant and Ask About the Wait for You

    Yelp’s New AI Will Call a Restaurant and Ask About the Wait for You

    The short version: On Tuesday, October 20, 2026, Yelp announced more than 30 new features in its fall update. The one most people will notice is called Check for Me. You tap a button in the Yelp app, and an AI voice phones the restaurant for you, asks how long the wait is, and texts you the answer within minutes. No calling, no being put on hold, no shouting your party size over a noisy dining room. It’s coming to iPhones in the U.S. first and will roll out “in the coming months,” so don’t panic if you don’t see it yet. Here’s how it works, what else changed in Yelp, and what to keep in mind.

    How Check for Me works

    Yelp says the whole thing takes just a few taps:

    1. Open a restaurant’s page in the Yelp app (or ask Yelp Assistant, the app’s built-in AI chatbot, about a restaurant).
    2. Tap Check wait time now.
    3. Type in how many people are in your group and your phone number.
    4. Yelp’s AI phone agent calls the restaurant and asks about the current wait.
    5. You get a text message with the answer, usually within a few minutes, according to Yelp.

    There are a few limits. The button only shows up at “eligible” restaurants where walk-ins are possible, so places that only take reservations won’t have it. It’s starting on iPhone in the United States. Yelp hasn’t said when Android phones will get it.

    What else is new in Yelp

    Check for Me is part of a bigger push to turn Yelp from a place to read reviews into an assistant that gets things done for you. Here are the changes most likely to matter to regular folks:

    • Book a table without leaving the chat. Yelp Assistant can now make reservations at tens of thousands of restaurants that use OpenTable, a popular booking service (on iPhone for now). You can also join a restaurant’s Yelp Waitlist from the chat and get an update when your table is ready.
    • Scheduled reminders. A new feature called Keep Me Posted lets you ask for things like “send me date night ideas every Wednesday” or “dog-friendly hikes once a month.” They arrive as phone notifications. You can change or turn them off in your Yelp settings, or just ask the assistant.
    • It remembers you. The assistant now uses your Yelp activity, including past chats and reviews you’ve written, to suggest things you might like. Yelp says it will soon use your searches and saved places too. Your recent chats are saved, with 30 days of history now and full history coming in the next few weeks.
    • It checks the weather. If rain is coming, it might steer you away from that patio-only spot.
    • My Lists. You can save places in a “Been Here” list and a “Want to Go” list, rank them, add notes, and share them with friends by text. Starting next month, Yelp will send weekly suggestions based on your lists.

    Yelp isn’t the only one making calls for you

    AI that phones businesses on your behalf is quickly becoming normal. TechCrunch points out that in the last few weeks, new personal assistant apps, including Meta’s Muse, have started calling businesses to handle customer support and book appointments. (We explained what Meta’s Muse does when it was announced.)

    Google has been doing a version of this for a while. Its Let Google Call feature, which launched in November 2025, phones nearby stores to check whether something is in stock and what it costs, then sends you a summary by text or email. Google’s help pages say its automated calls can also check restaurant wait times and book some appointments. Those calls are available in every U.S. state except Indiana, Louisiana, Minnesota, Montana, and Nebraska, and Google has said its caller tells the business it’s an automated system.

    The catch: robots calling robots

    Here’s the funny part. More and more restaurants are using AI to answer their phones, too. Yelp itself started selling an AI phone answering service to restaurants and other businesses last year. So your AI might end up chatting with the restaurant’s AI. TechCrunch notes that this kind of robot-to-robot conversation could break down. Picture two very polite answering machines saying “Sorry, I didn’t catch that” to each other until the end of time.

    And people still want people. A survey released the same day by the phone company Consumer Cellular found that 97% of its customers surveyed, all adults 50 and older, agreed there should always be an option to speak with a human. An AI that saves you a phone call is handy. An AI that stands between you and a real person when something goes wrong is a lot less fun.

    What this means for you

    • Don’t worry if you don’t see it yet. Check for Me is rolling out slowly, starts on iPhone, and only appears at restaurants that take walk-ins. Keep your Yelp app updated.
    • A wait time is a snapshot. “Twenty minutes” can turn into forty if a big group walks in right before you. Use the answer to decide where to go, not as a promise.
    • Call yourself for anything important. Food allergies, a wheelchair-accessible table, a big birthday group, or a special request are worth a quick conversation with a real person.
    • You’re giving Yelp your phone number. That’s how the answer gets texted to you. Only share it if you’re comfortable with that.
    • Know what the assistant remembers. Yelp’s assistant now uses your past chats and reviews to personalize suggestions. If that feels like too much, keep personal details out of your chats.
    • Watch for copycat scams. As AI calls and texts become common, scammers will copy them. If a text claiming to be from a restaurant or app asks you to pay, share a code, or tap a strange link, don’t. Open the app yourself instead.

    Try this next

    Curious how far these AI helpers can go? Read about why big websites are blocking AI agents, and how Google’s new Universal Cart watches prices for you. And if a text about your “reservation” looks fishy, here’s how to ask a free AI whether a text is a scam.

    Common questions

    What is Yelp’s Check for Me?

    Check for Me is a new Yelp feature, announced October 20, 2026, that uses an AI phone agent to call a restaurant for you and ask about the current wait. You tap Check wait time now, enter your party size and phone number, and Yelp texts you the answer, usually within minutes.

    When can I use Check for Me?

    Yelp says it is rolling out in the coming months in the U.S. on iPhone, at eligible restaurants where walk-ins are possible. Yelp hasn’t announced a date for Android phones.

    Can Yelp book a restaurant table for me?

    Yes. Yelp Assistant, the app’s AI chatbot, can now book tables at tens of thousands of restaurants that use OpenTable (on iPhone for now), and you can join a restaurant’s Yelp Waitlist right from the chat.

    Can Google call restaurants for me too?

    Yes, in many places. Google’s automated calling can check restaurant wait times, book some appointments, and check whether nearby stores have an item in stock. Google says it’s available in every U.S. state except Indiana, Louisiana, Minnesota, Montana, and Nebraska.

    Want tech that saves you time?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources

  • Google’s AI Search Mixed Up a Doctor with a Criminal. German Courts Say Google Is Responsible.

    Google’s AI Search Mixed Up a Doctor with a Criminal. German Courts Say Google Is Responsible.

    The short version: Google’s AI search told people that a respected German doctor had harassed his secretary, been fired, and been convicted in court. None of it was true. The AI had mixed him up with a different man. When Google refused to take it down, the doctor went to court, and on September 24, 2026, a court in Aachen, Germany, ordered Google to stop. It’s the second German court this year to say the same thing: when Google’s AI writes something, Google is responsible for it. The tech site The Decoder pulled the story together on Sunday, October 18. Here’s what happened, in plain English, and what it means for the AI answers you see every day.

    First, what are AI Overviews and AI Mode?

    When you search on Google these days, you often get a short AI-written answer at the top of the page, before the regular list of links. Google calls that an AI Overview. AI Mode is a newer option that works more like a chatbot: you ask a question and it writes you a longer answer, pulling from lots of websites at once.

    It’s handy. But the AI doesn’t copy one website word for word. It reads several, mixes the pieces together, and writes something brand new. Most of the time that works fine. Sometimes it stitches the wrong pieces together, like a friend who mixes up two people with the same haircut and tells a very confident story about the wrong one.

    What happened to the doctor

    The doctor had worked as a chief physician and professor at a German hospital from 2010 until late June 2026. On August 28, a colleague warned him about something strange. When people searched his name along with words like “suspension,” Google’s AI Mode said he had pushed himself on his secretary during a work trip and tried to kiss her, that he had been suspended and fired, that a court had given him a suspended sentence for sexual assault, and that he had walked away with a six-figure payout.

    Every bit of it was false. According to the German legal news site beck-aktuell, there was never a criminal complaint against him, no trial, and no conviction. The AI had apparently taken a real case involving a different former senior employee at the same hospital and pinned it on him. Even the news articles Google linked as its sources never named him.

    • He asked Google to fix it. In late August he emailed Google, explained the mix-up, and asked for the answers to be removed.
    • Google said no. In mid-September, Google replied that it had reviewed the complaint and would not take any action. A formal letter from his lawyers the same day didn’t change anything, and the false answers stayed up.
    • The court stepped in. On September 24, the Aachen Regional Court banned Google from spreading the three false versions of the story. Because the claims could badly hurt his career, the court acted as an emergency and didn’t wait to hear Google’s side first.
    • The price of ignoring it: If Google breaks the order, it faces a fine of up to 250,000 euros each time.

    This is a temporary emergency order, not a final ruling. Google can still challenge it.

    This isn’t the first time

    Back in late May 2026, a court in Munich ruled in a similar case. There, Google’s AI had falsely linked two Munich publishers to fraud schemes. The Munich judges said Google’s AI Overviews aren’t just a list of other people’s websites. They’re Google’s own statements, so Google has to answer for them.

    Google appealed that ruling, but according to The Decoder, the case quietly ended in a settlement before a higher court could weigh in. Nobody has said whether money changed hands. Now the Aachen court has used the very same reasoning, which suggests German courts are starting to agree on how to handle this.

    Why the courts said Google is responsible

    A search engine is usually seen as a go-between: it points you to what other people wrote. That’s a big reason search engines usually aren’t blamed for what’s on other websites.

    The German courts see AI answers differently. The Munich court said the AI makes “independent, new” statements by combining information from different websites. In other words, the AI isn’t pointing you to someone else’s words. It’s writing its own. And the Aachen judges said Google could have checked. They even noted that if Google had shown its own AI the doctor’s complaint, the AI would likely have figured out the claims were wrong.

    The doctor’s lawyer, Carsten Brennecke, put it bluntly (translated from German): “Whoever lets an AI make claims about people is liable for them. Google can’t hide behind an algorithm.”

    Does this affect people outside Germany?

    Not directly. These are German court decisions, and they only apply in Germany. Laws in other countries, including the United States, are different, and it isn’t clear yet how courts elsewhere will handle mistakes made by AI search.

    But the lesson travels well: AI search answers can sound completely sure of themselves and still be flat wrong, especially about real people. Google says so itself. Its own help page warns that AI Overviews “can and will make mistakes.”

    What this means for you

    • Treat AI answers about people as a first guess, not the truth. That goes double for anything about crimes, health, money, or someone’s job. AI is most likely to slip up when two people or businesses have similar names.
    • Click the sources. Under the AI answer, Google shows links to the websites it used. Open one or two and make sure they really say what the AI says. In the doctor’s case, the linked articles never even mentioned him.
    • Don’t make big decisions from a summary. Hiring a contractor, picking a doctor, or checking out a new date? Read the actual reviews, news stories, or official records, not just the AI’s short version.
    • Report wrong answers. At the bottom of an AI Overview, tap the thumbs-down icon, then Report a problem. You can also use the three-dot menu at the top of the overview. Google says this feedback helps it improve.
    • Look yourself up once in a while. Search your own name and your business’s name. If an AI answer says something false and harmful about you, report it, and for serious problems Google has a separate legal request form (search for “Report Content On Google”). It asks whether the problem is AI-made content. If it’s hurting your job or reputation, talking to a lawyer may be worth it.

    Try this next

    Want a second opinion you can trust a little more? Here’s how to ask a free AI chatbot whether a text is a scam, and the right way to double-check what it tells you. And if you use Google’s Gemini chatbot, here’s what its “Auto” setting means for free users.

    Common questions

    What happened with Google’s AI and the German doctor?

    Google’s AI Mode search falsely said a German doctor had harassed his secretary, been fired, and been convicted. It had mixed him up with a different man. Google refused to remove the answers, so on September 24, 2026, a court in Aachen ordered Google to stop spreading them, with fines of up to 250,000 euros per violation.

    Is Google responsible for what its AI search says?

    Two German courts, in Munich and Aachen, have said yes, because the AI writes new statements instead of just pointing to other websites. The Aachen order is temporary and not final, and these decisions only apply in Germany. Courts in other countries, including the U.S., haven’t settled the question.

    How do I report a wrong AI Overview on Google?

    At the bottom of the AI Overview, tap the thumbs-down icon, then choose Report a problem and pick the reason. You can also use the three-dot menu at the top of the overview. For legal problems, like false and damaging claims about you, Google has a separate legal removal request form.

    Can I trust Google’s AI Overviews?

    Use them as a quick starting point, not the final word. Google’s own help page says AI Overviews can and will make mistakes. For anything important, click the source links and check that they really say what the AI says.

    Want tech that’s genuinely helpful?

    We rounded up 27 unusual, well-reviewed gadgets that use AI to do something genuinely useful, explained in plain English.

    See our Tech Picks


    Sources