Google expands on‑device AI scam detection to 2 billion messages a

Google announced that its on‑device AI Scam Detection feature in the Messages app will now flag roughly 2 billion suspicious texts each month, a rollout that coincides with the upcoming Android 16 launch [1]. The move aims to curb the $470 million loss from text‑based scams reported last year and demonstrates Google’s push to embed privacy‑preserving AI across its mobile ecosystem.

On‑device AI tackles text scams

Google’s Scam Detection, introduced earlier this year, now covers a broader set of fraud types—including crypto, gift‑card, and technical‑support scams—while continuing to run entirely on the user’s device, meaning no message content is sent to Google’s servers [1]. Dave Kleidermacher, VP of Android security, highlighted the “epidemic” scale of financial scams, noting that the AI can observe longer, more sophisticated conversations (often called “pig‑butchering” scams) to spot malicious intent before users click harmful links [1]. Because the analysis happens locally, the system preserves user privacy while delivering real‑time alerts that let recipients report or block suspicious senders directly from the chat overlay.

Broader implications and next steps

The 2 billion‑monthly detection figure underscores the sheer volume of potentially fraudulent texts that Android devices now scrutinize, a scale not previously disclosed by Google. While other firms—such as Meta’s pop‑up payment warnings and O2’s “AI Granny” call‑bot—have introduced similar safeguards, Google’s approach is distinguished by its on‑device execution, which avoids the data‑centralization concerns that have plagued cloud‑based anti‑phishing tools [1]. The company also hinted at an upcoming expansion into phone‑call screening, though the feature remains in early testing and is not yet widely deployed [1]. If successful, this could extend AI‑driven protection beyond SMS to voice channels, potentially reshaping the broader mobile security landscape.

What to watch

• Android 16 rollout – Full deployment of the enhanced Scam Detection is tied to the Android 16 update slated for release next week.
• Call‑screening pilot – Google’s early‑phase testing of AI‑based scam detection for voice calls may indicate a broader security suite in future OS versions.
• Competitor responses – Watch for updates from Meta, O2, and cybersecurity firms that may introduce comparable on‑device AI safeguards to compete for user trust.

By embedding AI directly on smartphones, Google is positioning its security stack as a privacy‑first alternative to cloud‑centric solutions, while the sheer volume of flagged messages suggests a tangible impact on the fight against mobile fraud. The effectiveness of the upcoming call‑screening pilot and the speed of Android 16 adoption will determine whether this strategy reshapes the market or remains a niche feature.