Google has launched AI-based defenses against messaging scams and call forwarding abuse on Android devices. Existing Phone and Messages systems flag over 5 million suspicious calls and messages monthly. The new defenses include cloud-based analysis of complex calls using Gemini, on-device link checks using Gemini Nano, and clearer warnings before opening links from unsaved numbers.
What changed in the protection systems
The new defenses operate on two levels. On-device checks using Gemini Nano scan links in messages and identify suspicious patterns without sending information out. Cloud-based analysis using Gemini examines complex calls where the scammer tries to create pressure or urgency. Cloud-based defenses are off by default, and the user must activate them manually.
The new warnings appear before opening links from unsaved numbers, even if the link itself has not been identified as malicious. This means any message from an unknown number containing a link will display a warning, regardless of content.
Why this matters for businesses communicating with customers
Businesses using automated messages, bots, or CRM systems that send messages to customers need to understand that their messages now pass through filters looking for scam patterns. A message sent automatically right after a purchase or registration, containing a verification or tracking link, and coming from a number the customer has not saved, meets three criteria that anti-scam systems look for.
The problem is more acute when dealing with changing numbers or sending from a number that is not the business’s official number. If you use a bulk messaging service or WhatsApp Business API, ensure the sending number is clearly identified as belonging to the business, and that customers receive clear instruction to save it.
What to do this week
Check the automated messages you send to customers. If they contain links, consider splitting the communication: a first message without a link that introduces the business and asks the customer to save the number, and a second message with the link only after the customer has responded or confirmed. If you use phrases like “act now”, “limited offer”, or “verify your account immediately”, replace them with more neutral language.
If you are building a new automated communication system, plan ahead how your number will appear to the customer, how it will be saved, and how the first messages will look. A system built correctly from the start will not need to fix damage after customers start complaining that messages are not arriving. If you need help planning automated communication that works under modern security rules, see managed automation for businesses.
Sources
- Android gets AI defenses against messaging scams, fake calls · BleepingComputer
Frequently asked questions
Are WhatsApp Business messages affected by Google's new defenses?
The new defenses operate in Google's Phone and Messages apps, not in WhatsApp. But WhatsApp already runs its own scam detection systems, and businesses sending automated messages with links from unsaved numbers encounter similar problems there too.
How do I know if my messages are being blocked or flagged as suspicious?
There is no direct way to know, but if you see a drop in message open rates, an increase in complaints that customers did not receive a message, or customers reporting warnings, that is a sign the system is flagging you. Check whether your messages contain links, come from changing numbers, or use urgent language.