Google Builds a Digital Handshake to Stop AI Voice Scams

The News

Google has switched on a new defence against one of the fastest-growing forms of phone fraud: AI-driven voice impersonation. On Tuesday, 2 June 2026, it began rolling out fake call detection inside the Phone by Google app, designed to flag calls where a scammer poses as someone you trust.

The system works as what Google calls a "digital handshake between devices". When a genuine contact who also uses Phone by Google rings you, their handset sends a silent confirmation signal verifying the call is real. If a fraudster spoofs that number, the signal is missing, and your phone quietly pings the contact's actual device to check. If the real device reports it is not making a call, you get an on-screen warning to hang up.

The feature is on by default, runs automatically, and reaches Android 12 and newer. It is rolling out globally this month, starting with Pixel phones, and is built on Rich Communication Services (RCS), so other apps and carriers can adopt it.

Why It Matters

Caller ID has been broken for years. Number spoofing is trivial, and cheap voice cloning has turned a nuisance into a financial weapon. A scammer no longer needs to sound vaguely official; with seconds of sampled audio they can mimic a relative in distress or a manager demanding an urgent transfer.

Google's approach is notable because it does not try to detect the fake voice, a cat-and-mouse game generative models keep winning. Instead it verifies the channel: is this call really coming from the device it claims to? That is a structural fix rather than a detection arms race. The last comparable effort was the STIR/SHAKEN framework US carriers rolled out from 2021, which authenticated networks but never reached the handset-to-handset layer Google now targets.

Building on RCS matters too. By making verification an open standard rather than a Pixel-only perk, Google is treating caller trust as shared infrastructure other manufacturers and operators can plug into.

Indian Angle

Nowhere is this more relevant than India. The country runs on Android, which powers the vast majority of its smartphones, and it has become a global epicentre of voice-led fraud. The government's cyber-crime machinery, including the I4C unit under the Home Ministry and the 1930 fraud helpline, has spent two years warning about "digital arrest" scams, where callers impersonate police, couriers, or tax officials to extract money. Deepfaked voices have made those scripts far more convincing.

For Indian banks, fintechs, and the RBI, the timing matters. UPI-era fraud-awareness campaigns have raised the alarm, but the weak link is the call itself. A handset-level verification layer could blunt the most common attack vector before a victim ever opens a banking app, complementing MeitY's earlier push on deepfake labelling.

There is a catch. The rollout starts with Pixel, a phone with negligible share in India, where Xiaomi, Samsung, Vivo, and Realme dominate. The real test is how fast Indian-market handsets and RCS-heavy carriers such as Jio and Airtel adopt the standard. Until they do, the most exposed users will be protected last.

FAQ

When does this take effect?

Google began the global rollout on 2 June 2026, starting with Pixel devices and expanding to other Android 12 and newer phones through the month. Availability on non-Pixel handsets will depend on manufacturer and carrier updates.

How is this different from spam call blocking?

Spam blockers screen unknown numbers. This feature verifies known contacts, confirming a call claiming to come from a trusted person's device genuinely originates there, which is precisely the gap deepfake impersonation exploits.

Will it work on my phone in India?

It requires Android 12 or later and the Phone by Google app. Pixel users get it first; most Indian-brand phones will need to wait for the feature to arrive via the wider RCS ecosystem and carrier support.

Does it stop AI voice cloning?

Not directly. It does not analyse the voice at all. It checks whether the call really comes from the device it claims to, so a cloned voice on a spoofed number is flagged however convincing it sounds.

This story was reported by TechCrunch. Read the full original coverage at TechCrunch.