Android Adds More AI Protection Against Scam Texts and Fake Bank Calls
Google is adding Gemini-based scam checks to Messages, blocking apps from silently forwarding your calls, and expanding verified bank calls. Some of it is opt-in, and several parts arrive only 'later this year'.
Google is rolling out a batch of AI-powered scam defences for Android, aimed at the tricks that actually fool people: suspicious texts, dodgy links and phone calls that pretend to come from your bank. BleepingComputer reports the features are going to “eligible devices,” without saying which ones qualify.
In Google Messages, a small Gemini Nano model running on your phone checks unfamiliar links using real-time web signals and warns you before you open a link from a number that is not in your contacts. For harder cases, such as long chats with unknown senders, a cloud-based Gemini check looks for signs of fraud. Google says that runs in an encrypted “Private AI Compute” environment, but it is off by default, so you have to switch it on yourself. Scam detection in chat notifications, previously a Pixel feature, is due to reach “hundreds of millions” more people through Android partners later this year.
Two changes target specific tricks. Apps can no longer quietly switch on call forwarding in the background, and you get a warning before a forwarding code you typed in takes effect. (Call forwarding abuse lets a scammer receive the verification call meant for you.) And “verified bank calls” check that a call really was started by your bank’s call centre; if the check fails, the call ends automatically. Google says it has blocked more than 3 million fraudulent calls for two Brazilian banks alone since June, and that the system is expanding to more institutions, with Revolut named.
The idea behind it: scammers have always relied on pressure and speed. Putting a small AI model on the phone itself, instead of sending your messages to a server, lets Google look at links and chats as they arrive while keeping most of the checking private. The honest limit is that these are filters, not guarantees, and anything that needs the cloud check requires your consent.
What this means for you: if you use an Android phone, open Messages settings and look for the scam protection options. Decide whether the opt-in cloud check is worth it to you; it is a privacy trade-off, and a reasonable one either way. Whatever the software does, the old rule still holds: your bank will not ask for codes over the phone, and you can always hang up and call back on the number printed on your card. Do not wait for a rollout to teach this to older relatives.
Sources
Anthropic's Survey Finds a Huge Gap Between How the Public and Its Own Staff See AI's Future
Anthropic surveyed about 11,000 US adults and 1,334 of its own staff. They agree on what AI does today, but not on how fast it will improve or how much of the work it will take over.
