The bank itself becomes the source of truth.
As phone scammers drain nearly a billion dollars from the global economy each year, Google is responding not with warnings but with architecture — rebuilding the trust layer of Android itself. The 2026 update treats the smartphone as a contested space where identity can be spoofed, devices can be stolen, and AI can become a liability as easily as a tool. In answering these threats, Android is quietly redefining what it means for a device to be on your side.
- Phone scammers spoofing bank numbers cost the world $980 million a year — and Android is now building a real-time verification system that cuts the call if your bank's own app can't confirm it's calling.
- Theft protection is escalating: lost devices will require biometric unlock, hide Quick Settings, and block new wireless connections, with these defenses switching on by default across new and reset devices.
- Live Threat Detection will watch apps in real time for behaviors like SMS forwarding and accessibility abuse — the quiet, technical moves that turn a phone against its owner.
- Five countries are receiving automatic activation of Remote Lock and Theft Detection Lock for any device running Android 10 or newer, signaling a shift from opt-in security to enforced baseline protection.
- AISeal introduces hardware-backed isolation for AI workloads, ensuring that the ambient data powering smart features never leaks into the broader system — privacy-preserving AI as infrastructure, not afterthought.
Phone scammers have long relied on a simple trick: spoof a bank's number, call a customer, and convince them to hand over money or credentials. The scheme costs the world roughly $980 million a year, and Google has decided Android's 2026 update will confront it directly.
The answer is verified financial calls — a system built with real banks like Revolut, Itaú, and Nubank. When a call arrives claiming to be from your bank, Android checks whether the bank's own app confirms an outgoing call is actually in progress. If it doesn't, the connection is terminated. Banks can also flag certain numbers as inbound-only, automatically cutting off any call that originates from them. The rollout begins in the coming weeks on Android 11 and newer.
Android 17 also introduces Live Threat Detection, an on-device AI that monitors apps in real time for suspicious behavior — SMS forwarding, accessibility overlay abuse — the quiet tactics used to steal data or hijack a device. Chrome on Android gains pre-download malware scanning for APK files, and Advanced Protection mode now strips accessibility access from apps that aren't labeled as accessibility tools, among other hardening measures.
Theft protections are expanding meaningfully. Marking a phone as lost through Find Hub will require biometric authentication to unlock it, hide Quick Settings, and block new wireless connections. These protections will be on by default for all new Android 17 devices and any phone that's reset or upgraded. In Argentina, Chile, Colombia, Mexico, and the UK, Remote Lock and Theft Detection Lock will activate automatically on any device running Android 10 or newer. PIN guess limits and longer wait times between failed attempts add another layer of friction for would-be thieves.
Location and contact permissions are also being tightened — a one-time location button, a visible indicator when apps have recently accessed your location, and a redesigned contact picker that grants access to individual contacts rather than entire address books.
The most forward-looking addition is AISeal, a hardware-backed isolation layer for AI workloads. As AI features become standard on smartphones and require access to ambient background data, AISeal ensures that processing happens in a verified, contained environment — separate from the broader system. It is Google's clearest signal yet that the privacy question surrounding on-device AI is not being deferred, but designed around.
Phone scammers have found a reliable playbook: spoof a bank's phone number, call someone claiming to be from their financial institution, and convince them to hand over money or credentials. The scheme works often enough that it costs the world nearly a billion dollars a year. Europol estimates the annual toll at $980 million, and Google has decided that Android's 2026 update will take a direct swing at the problem.
The centerpiece is a system called verified financial calls, built in partnership with actual banks. Here's how it works: when a call arrives claiming to be from your bank, Android checks the bank's official app on your phone to see if the bank is actually placing a call right now. If the app says no, the system terminates the connection. Banks can also mark certain numbers as inbound-only, which means any outgoing call from those numbers gets cut off automatically. Eugene Liderman, who directs Android's security and privacy team, said the rollout begins in the coming weeks with Revolut, Itaú, and Nubank on Android 11 and newer devices, expanding to more banks later in the year.
Beyond phone calls, Android 17 is adding what the company calls Live Threat Detection—an on-device AI system that watches for suspicious app behavior in real time. The new version will flag apps that try to forward text messages or abuse accessibility overlays, two common tactics for stealing data or taking control of a device. Chrome on Android is getting its own safeguard: when Safe Browsing is enabled, the browser will scan APK files for malware before letting you download them. Advanced Protection, a hardening mode introduced last year, now removes accessibility service access from apps that aren't labeled as accessibility tools, disables device-to-device unlocking, blocks Chrome WebGPU support, and adds scam detection for chat notifications.
Theft protection is expanding significantly. On Android 17, if you mark your phone as lost through Find Hub, the system will require your fingerprint or face to unlock it—a crucial defense against thieves who may have watched you enter your PIN. Marking a device as lost also hides Quick Settings and prevents new Wi-Fi and Bluetooth connections. Following a pilot in Brazil, these protections will be on by default for all new Android 17 devices and any phone that's reset or upgraded. In Argentina, Chile, Colombia, Mexico, and the UK, the protections go further: any device running Android 10 or newer will automatically activate Remote Lock and Theft Detection Lock. Android 17 also limits the number of PIN guesses allowed and increases the wait time between failed attempts. On Android 12 and later, you can access your IMEI from the lock screen to help law enforcement and carriers verify ownership.
The update also tightens control over location and contact data. Android 17 adds a one-time location button that lets you share your precise location for a specific task without granting permanent access, and a top-of-screen indicator shows when apps have recently accessed your location. A redesigned contact picker lets developers request access to individual contacts and specific fields rather than your entire address book, and those permissions are temporary by default.
Perhaps most forward-looking is AISeal, a new hardware-backed isolation layer for AI workloads that pairs with Android's existing Private Compute Core and Private AI Compute systems. It creates a verified environment for processing ambient data—the kind of background information that AI features increasingly need to work well—without exposing that data to the broader system. As AI becomes standard on smartphones, the question of where and how that data gets processed has become urgent. AISeal is Google's answer: keep it isolated, keep it private, keep it on the device.
Notable Quotes
We'll begin rolling this out on Android 11+ devices with Revolut, Itaú, and Nubank in the coming weeks, before expanding it to more banks later this year.— Eugene Liderman, Director of Android Security & Privacy Team
The Hearth Conversation Another angle on the story
Why does a bank call scam work so well that it costs nearly a billion dollars a year?
Because the person on the phone sounds official, the caller ID says it's the bank, and most people trust what they see on their screen. By the time you realize it's not actually your bank, you've already given them something they can use.
And verified financial calls stops that how, exactly?
It checks with the actual bank app on your phone. If the bank app isn't making a call, then no call from that number is real. It's simple—the bank itself becomes the source of truth.
What about people who don't have the bank app installed?
That's the gap. The system only works if you have the app. But the rollout starts with major banks, and most people who use those banks probably have the app already.
The theft protections seem like they're trying to solve a different problem—someone who physically has your phone.
Exactly. If your phone is stolen, the thief has your device but not your biometric data. They can't guess your PIN forever because the system locks them out. And if you mark it as lost remotely, it becomes nearly useless to them.
What's AISeal doing that's different from what Android already has?
It's hardware-backed isolation. When your phone is processing ambient data for AI features, AISeal creates a locked box that even Google can't easily open. It's saying: we're going to use AI, but we're going to do it in a way that keeps your data away from everything else on the phone.
Does that mean the AI features work better, or just more privately?
Both, potentially. If the AI can process data without worrying about other apps stealing it, it can be more aggressive about what it learns. And you get privacy as a side effect.