Android phones will soon be able to detect spoofed calls and impersonation scams

The buried angle here is what this does to the carrier and third-party security app market. If fraud detection moves to the OS layer, companies like Hiya, First Orion, and T-Mobile's Scam Shield are looking at a commoditization problem, not a partnership opportunity.

This fits a pattern Modelwire has been tracking across the week: AI inference moving closer to the hardware and OS layer, away from cloud or app-level intermediaries. The Nvidia RTX Spark coverage (The Decoder, June 1) framed local inference as a structural shift in who controls the edge AI stack. Google's move here is the same logic applied defensively at the OS level rather than offensively for productivity workloads. It also sits in direct tension with the Meta Instagram account takeover story from June 1, where an AI system's compliance-oriented design became the attack surface. Google is betting that on-device ML can close the gap that social engineering exploits, but the Meta incident is a reminder that the model's behavior under adversarial prompting matters as much as its deployment layer.

Watch whether Apple announces a comparable on-device call authentication feature at WWDC 2026. If Apple ships nothing by end of June, that signals Google is using this as a differentiation point rather than responding to a shared industry standard.