Google Messages now uses on-device AI to spot text scams

Modern text scammers don't just blast sketchy links anymore. Instead, they start with a polite wrong-number text, chat casually to build trust, and only drop the trap days later.

To fight back, Google is rolling out an on-device AI scam detection feature inside Google Messages that monitors conversations in real time for manipulative tactics.

Spotting scams mid-chat

Traditional spam filters are built for the front door. They check the sender and look for obvious keywords before the message even lands in your inbox, but they are practically useless against modern social engineering scams.

According to Alberto Pastor Nieto, senior product manager for Google Messages, conversational scams begin innocently before pivoting into dangerous schemes. Google's new AI models are designed specifically to follow SMS, MMS, and RCS chats as they unfold.

When the system flags a suspicious pattern, it shows an instant warning banner letting you dismiss the alert or immediately block and report the sender.

To protect user privacy, Google says all message processing happens entirely on-device, meaning conversations never leave your phone. The feature is enabled by default, monitors only unsaved contacts, and can be toggled off in settings. It is launching first in English across the US, UK, and Canada, with other regions coming soon.

Pixel 9 gets multicam and satellite upgrades

Alongside the security update, Android is adding live location sharing with trusted contacts inside the Find My app, complete with regular reminders so you don't forget who can see you.

Pixel 9 owners are also getting several hardware-focused upgrades. The phones can now connect to GoPro cameras or other Pixel devices over Bluetooth and Wi-Fi for multicamera live streaming on YouTube, Instagram, and TikTok, mirroring Apple's multicam setups.

Pixel Studio can now generate stylized people and stickers, intentionally avoiding photorealism to prevent deepfake misuse. Pixel 9 devices are also adding satellite messaging support across both Verizon and T-Mobile networks, while the Recorder app can now import audio from older Pixel devices and transcribe it automatically.

Why it matters

We think shifting scam protection from initial inbox triage to ongoing, real-time conversation analysis is a remarkably smart move. Pig butchering and romance scams succeed precisely because early messages look harmless enough to breeze past standard spam filters.

Pulling this off entirely on-device is the only acceptable way to do it. It protects sensitive personal chats from cloud exposure while giving everyday users an automatic safety rail against sophisticated social manipulation.

That said, the real test will be false alarms. If a lightweight on-device model misinterprets an awkward conversation with a new acquaintance or flags legitimate financial chats too aggressively, alert fatigue will set in fast.

Even so, while Apple has added RCS support and cleaner spam sorting, Google is pushing deeper into contextual, proactive protection. It sets a new standard for what default messaging apps should do to protect their users.