Google plans to roll out a new, opt-in tier of Scam Detection in Google Messages that uses Private AI Compute and cloud Gemini models, alongside on-device Gemini Nano link scanning, according to an announcement made by the company. The features are set to begin rolling out in the “coming weeks,” and both require users to actively turn them on.

What is actually changing in Google Messages

The headline item is officially called Scam Detection enhanced with Private AI Compute. Rather than relying only on existing signals, Google says it will use its most capable cloud Gemini models to examine the full context of a conversation and attempt to understand a scammer’s intent — including “slow-burn” scams that unfold gradually and are harder to catch.

The analysis applies to conversations with unknown senders. To address the obvious privacy problem of sending message content to the cloud, Google routes the work through Private AI Compute, which encrypts and processes data in a secure dedicated space. The feature is opt-in, so nothing runs unless the user enables it once it goes live.

Which devices get it

The new Scam Detection level will be available on:

  • Pixel 6 and up, through the Pixel 11 series
  • Galaxy S24 and up
  • Galaxy Z Fold 6 and Flip 6 and up
  • Other Android AICore devices

That device list is notable because it defines how far the cloud-assisted detection reaches. Older or lower-tier Android hardware outside the AICore program is not covered by this announcement.

Link checking moves on-device with Gemini Nano

Google is also working on the link side of scam defense. Through Enhanced Protection, it will use on-device Gemini Nano to evaluate unfamiliar links and judge whether they may be scams. The rationale Google gives is specific: scammers frequently register brand-new websites for each new scam, so earlier checks may have had nothing to flag them with. On-device Gemini Nano scanning is meant to close that gap.

Enhanced Protection is also getting redesigned warnings when a scam link is detected. The new warnings add extra time and clearer context, giving users a moment to pause before tapping something harmful.

Smart link checking will be available on a narrower set of hardware: Pixel 10 and up, the Galaxy S26 series, and the Galaxy Z Fold 8 and Flip 8.

Why this matters for readers

Scam messages are a volume problem, not a novelty problem. A detection layer that only matches known bad senders or known bad links struggles against campaigns that rotate infrastructure and build trust slowly over many messages. Google’s answer here is two-pronged: cloud Gemini models for intent and context on the message side, and on-device Gemini Nano for freshly registered links that no reputation database has seen yet.

The trade-off is hardware and consent. Cloud-assisted intent analysis is limited to devices in the AICore program, and on-device link scanning is limited further to newer Pixel and Galaxy models. Both features are opt-in, which means the protection is only as widespread as users choose to make it.

What to watch next

The rollout window is described as the “coming weeks,” so availability will likely arrive in stages rather than all at once. Two things are worth tracking: whether the opt-in prompt explains Private AI Compute clearly enough for users to make an informed choice, and whether the redesigned Enhanced Protection warnings actually change behavior when a link looks suspicious. Until then, the practical step for eligible device owners is to watch for the setting after the update lands.