Google Is Adding New Tools to Catch Text Scams Before You Click

Google Messages will add a new "enhanced protection" switch in the coming weeks. You will have to turn it on yourself. It is meant to catch scams from people you do not know. The Verge

When it is on, Google's Gemini AI in the cloud reads the whole chat to judge what the stranger is trying to do. It does not just look for bad words. It watches how the sender pushes the talk forward over several messages.

Google says this cloud check helps the scam check already on your phone catch new and tricky attacks. Your phone is good at fast, private checks. The cloud can use a bigger AI model that learns new scam tricks faster. The trade-off is that some message data must leave your phone, even if it is encrypted. That is why it is opt-in.

Google says that data is encrypted and handled in a locked-down private area that even Google and others cannot see besides you.

A second check stays only on your phone. Enhanced protection uses Gemini Nano, a small AI that runs on the device, to study strange links before you tap them. It looks at live clues on the web to decide if a link is dangerous.

That matters because the text can look normal while the link leads to a brand-new fake site. Checking fresh clues at the moment you might click catches fast-changing scam sites better than an old block list.

OS-level controls

Android itself is getting new guards too. Android will stop apps from turning on call forwarding by themselves and will tell you if an app tries. Once call forwarding is on, your calls can be sent to a scammer's number. That lets them grab return calls and one-time codes. Blocking that path stops apps from doing it quietly.

Android will also hide private details automatically when you share your screen in Chrome and Android WebView, the tool apps use to show web pages. Fake support scams often need you to share your screen while you log in or pay. Hiding that information lowers what scammers can steal, even if you do not spot the scam yet.

Scam warnings that pop up in chat notifications started on Pixel phones last year. They are now coming to Honor, Motorola, Oppo, OnePlus, Realme, Tecno, Vivo, Xiaomi and more. Google plans to reach hundreds of millions more people later this year. That broadens cover because it can catch scam texts from many chat apps, not just Google's own.

Where this fits

These tools build on work from the last two years. Google Messages already uses AI to flag chat patterns tied to scams, including package delivery and job scams that are very common now. Google Google has said this AI scam check blocks billions of suspect messages.

Chrome's Enhanced Protection uses Gemini Nano for extra scam defense. Chrome on Android has also added AI warnings about spammy alerts.

Scammers are also using AI at scale. Google says scammers used Gemini AI to help write spam, with one group accused of sending 2.5 million messages to Android users in two weeks. Bloomberg Google filed a New York lawsuit against makers of software said to enable AI-powered phishing. Reuters

Regulators are pushing as well. EU consumer groups complained that Google, Meta and TikTok do not remove fake ads. India ordered removal of Google Firebase accounts tied to Android malware dressed up as bank apps. Singapore police ordered Apple and Google to stop scammers from posing as government agencies in messages.

The broader context here is a move from blocking known bad links and numbers to watching intent and behavior. Making new lures is cheap, so the plan assumes some will get through. The work then focuses on what harm can still be stopped after delivery, like a rerouted call or a stolen password on screen.

In my view, the opt-in part is worth flagging for families and work teams. The people who most need the cloud check are least likely to turn it on. Turning it on by default would help more at-risk users but would raise questions about chats leaving the phone. Google chose clear consent plus its locked-down processing claim. How many people turn it on, how often it flags normal chats from strangers, and how fast the link check feels will decide if this becomes normal protection. Over time, these kinds of guards tend to become defaults once the privacy design is trusted, and this is a step toward that.