Google Messages will scan chats with strangers in the cloud

Google Messages will get an opt-in mode that sends conversations with unknown senders to Google's servers, where Gemini artificial intelligence models look for scams. The company announced it on Oct. 14 on the Android security blog and said it will arrive "in the coming weeks." It gave no country-by-country dates, so it is unclear whether or when it will reach Mexico.

Today, Messages detects scams on the phone itself. The new "Enhanced protection" setting changes where message content is analyzed: if the user turns it on, the analysis moves to the cloud. Google says it is meant to catch slow-burn fraud, the kind that only shows itself after several messages.

What Google says about privacy

The cloud analysis runs on Private AI Compute, an environment Google describes as a secure, dedicated space. According to the company, the data is encrypted and out of anyone's reach, itself included. That is the company's own claim. Outlets covering the announcement cite no independent technical evaluation, and Google does not say how much of each chat is sent or how long it is kept.

Which devices are covered

According to Google, cloud detection will work on these devices:

• Pixel 6 and later.
• Galaxy S24 and later, and Galaxy Z Fold6 and Flip6 and later.
• Other Android phones with AICore, the component that runs AI models on the device.

The package adds other features. Gemini Nano, the small version of the model that runs on the phone, will check unfamiliar links on Pixel 10 and later, the Galaxy S26 series and "other select devices." Warnings about links sent from numbers outside the address book are also being redesigned.

According to Google, its system flags more than 5 million calls and messages a month that it considers fraudulent. The figure comes from the company, and nothing published verifies it externally.

The announcement notes that some features may not reach every country. Still pending is a list of markets from Google and an activation date.