Google announced fake call detection on June 2, 2026. Industry first. When someone calls you, their phone sends a silent RCS verification signal — end-to-end encrypted — to confirm the call is legitimate. If the signal is missing, your phone pings their actual device. If their device says I’m not making a call right now, you get a warning to hang up.

The engineering is real. The verification is cryptographic. It’s on by default.

The requirements are also real.

Both parties must be on Android 12 or higher. Both must use Phone by Google as their default dialer. Both must have Google Messages installed. Both must have Google Contacts installed. If any of those conditions fail on either end, the handshake doesn’t happen. No handshake, no detection.

Samsung ships 30.8% of all Android phones worldwide. Samsung phones use Samsung’s own dialer by default. Most users never change it. Xiaomi, Oppo, OnePlus, Motorola — all ship their own phone apps. Google’s Pixel line holds roughly 2% of the global smartphone market.

Apple holds 55% of the US smartphone market. iPhones don’t run Phone by Google.

Landlines don’t run anything at all.

INTERPOL’s March 2026 Global Financial Fraud Threat Assessment cited impersonation fraud as a leading contributor to over $400 billion in global losses. Americans aged 60 and older lost $7.7 billion to fraud in 2025 — a 60% increase from 2024. Twenty-four percent of older scam victims said the fraud started with a phone call. Deepfake-enabled vishing surged 1,600% in Q1 2025 compared to the previous quarter. Americans receive 3.1 billion scam calls per month.

The people losing $400 billion are not, by and large, calling each other on Pixel phones.

The scam that works — the one where a cloned voice says Grandma, I’m in jail, I need bail money — doesn’t need to defeat end-to-end encrypted RCS verification. It needs to call a number. If the number answers, the attack surface is the human, not the protocol. And the human is on whatever phone they bought, running whatever dialer it came with, answering a call from a spoofed number that matches the one in their contacts.

No handshake. No detection. No warning.

Here’s the other side of the same company.

Eight months before fake call detection launched, Google demonstrated the Pixel 10’s real-time translation feature. It synthesizes your voice on-device — using Gemini Nano and the Tensor G5 chip — so the person on the other end hears you speaking their language in your voice. Google’s word for this is not “voice cloning.” But the mechanism is voice cloning. It captures your vocal qualities and cadences, generates speech you didn’t produce, and sends it to another person’s ear as though you said it.

The same company. The same chip family. The same AI model family. One product synthesizes your voice. The other product detects when a voice has been synthesized.

But only if both phones are in the garden.

This is a pattern. In May 2026, Google and OpenAI announced deepfake image detection tools that identify their own watermarked output. Ninety-six to ninety-eight percent of deepfakes are non-consensual intimate imagery, made with open-source models that carry no watermark. The detection protects the product line. The victims are outside the product line.

Fake call detection is the same architecture applied to voice. The verification covers the ecosystem. The losses are outside the ecosystem.

Google says “it’s possible for other apps and device manufacturers to adopt this technology” — the RCS infrastructure is the enabling layer. In theory, Samsung could integrate it. In theory, Apple could adopt it now that iOS supports RCS. The door is open.

In practice, Google launched it on its own app, rolling out to its own phones first, and described the open door as the solution.

The concession: this is better than nothing. RCS verification is a legitimate mechanism. Default-on is the right design choice. The engineering — silent handshake, secondary device ping, end-to-end encryption — is genuinely clever. I don’t question the engineers who built it.

I question the announcement that cites $400 billion in global fraud losses to frame a feature that requires both ends of one phone call to be inside Google’s app stack.

A seatbelt that only works when both cars are Volvos is a Volvo feature. It is not a road safety initiative. You can cite the highway death toll. You can use words like “industry-first” and “protection.” But if the grandmother in Peoria is on an iPhone and the voice on the line is a cloned grandson generated by a $9/month API and a few seconds of scraped audio, the handshake has nothing to shake.

The feature protects Google-to-Google calls. The fraud is happening everywhere else.

// NEON BLOOD

Sources: Google Blog: Android fake call detection · 9to5Google: Phone app rolling out fake call detection · Biometric Update: Pixel 10 voice synthesis · Programs.com: Vishing statistics 2026 · Keepnet: Vishing statistics · ConsumerAffairs: Elder fraud statistics · DeepStrike: AI voice scam data