Imagine this message arrives while you are waiting for a parcel: “We couldn’t deliver your package. Pay a small redelivery fee here.” There is a link. Your phone shows no warning.
This is an invented example, but the decision is familiar: should the absence of an alert make the link feel safer? It should not be treated as verification. A scam detector can miss a message, and a particular protection might not cover that device, app or interaction in the first place.
Google’s October 14 Android announcement describes several new defenses, including an optional Enhanced protection mode in Google Messages. Google says this mode will add cloud-based analysis of complex conversations with unknown senders, alongside other protections. The idea is to look beyond a suspicious word or a previously identified bad link.
What did the AI actually check?
A separate, existing feature warns about scams in chat notifications. Google’s help page says it examines new message notifications from supported apps on selected Android devices. That is a specific input, rather than a promise to inspect everything you do on your phone.
The same page says detection may miss scams. It also explains that marking a message as “Not a scam” temporarily turns detection off for that conversation. A quiet conversation, therefore, does not necessarily mean the system has checked every message and approved it.
Think of two separate questions: Was this interaction covered? And, if it was, did the detector recognize a suspicious pattern? Neither question establishes that the person asking for payment really represents the delivery company.
For this chat-notification feature, Google says message content stays on the device unless you choose to share it. The new Google Messages mode has a different design: it adds cloud analysis through Google’s Private AI Compute. Those are descriptions of two different protections; a privacy statement about one should not be assumed to describe the other.
An announcement is not a setting on every phone
As of October 21, the rollout language matters. Google’s announcement puts Enhanced protection in Google Messages in the coming weeks and wider chat-notification support later this year. Its call-forwarding protection is scheduled to begin in early December. These features should not all be described as already available to every Android user.
For the existing notification feature, Google’s help page describes eligibility and lists supported countries and languages and suggests searching Settings for “Scam Detection.” Check the relevant feature on your own phone instead of relying on the announcement’s headline.
Back to the parcel. For a text claiming to concern a USPS delivery, the U.S. Postal Inspection Service advises against clicking suspicious links even when you are expecting a package. It recommends visiting USPS.com directly for tracking. Our example calls for the same habit: open a trusted route yourself and check the delivery there, instead of following the message’s instructions.
An AI warning can interrupt a rushed decision. When no warning appears, you still need a reason to trust the request.