Your phone says “spam” before you answer. How AI spots suspect calls

Ever wondered how your phone flags an unknown caller as suspected spam before you answer? Increasingly, artificial intelligence (AI) is helping make that call.

Telecom operators in India use AI and machine learning to study calling patterns and warn customers about suspected spam. The Telecom Regulatory Authority of India (TRAI) said in September that major operators had already deployed these systems. It has now made AI-based detection part of its anti-spam rules.

The scale is substantial. Operators received 721.79 billion incoming calls during April-June 2026, with AI systems flagging 22.99 billion as suspected spam, according to the Ministry of Communications. These systems also flagged 1.44 billion of 836.31 billion incoming SMS messages.

Reading the calling patterns

Airtel says its system uses information from its network to identify suspected spam calls and messages, displaying an “Airtel Warning: SPAM” label. It examines how often a number calls, how long calls last, usage patterns, suspicious web links and the IMEI, a device’s unique identification number. The company says it uses more than 250 derived signals to assess spam behaviour.

Vodafone Idea, or Vi, says its Voice Spam Detection system uses AI models, tools that scan the web and user feedback to identify suspicious callers in real time. Customers see a “Suspected SPAM” warning when a flagged number calls.

Screening callers and voices

Caller-ID app Truecaller adds another approach to its database of spam numbers reported by users. In 2024, it launched AI Call Scanner, which analyses a caller’s voice during a call to indicate whether it is human or AI-generated, according to the company.

Users can activate the scanner during a suspicious call. Truecaller says the feature is currently available in India only to beta users, who test features before a wider release.

Its AI-powered Assistant can also answer calls and ask callers why they are calling, helping users decide whether to pick up. Truecaller says it uses machine learning and speech-to-text technology, which converts speech into written words. Most processing happens in the cloud.

Google’s Phone app offers Scam Detection on supported Pixel devices in India. Google says it uses Gemini Nano, an AI model running on the phone, to flag potential scams during calls. Google says processing stays on the device, and no conversation audio or transcript is stored or sent to its servers.

“Scam Detection is not 100% accurate and scammers constantly change their tactics,” Google says on its support page.

Google has separately introduced fake call detection on Android to flag calls that falsely appear to come from a known contact. It checks whether the call comes from that contact’s device when both people use Phone by Google, helping guard against impersonation scams involving cloned voices.

From warnings to action

Under TRAI’s amended Telecom Commercial Communications Customer Preference Regulations, 2018, announced in September, operators must identify numbers likely to send unwanted commercial communications and share details with other operators.

If five or more numbers linked to a sender are flagged within 10 days, operators must investigate and begin action in stages. This can include rechecking customer identity documents, physical verification, blocking outgoing services and disconnecting telecom connections for repeated violations.

The rules also cover application-to-person (A2P) calls, placed automatically by software. These include autodialled calls, robocalls and calls using pre-recorded or artificial voices. Businesses must declare their use and the numbers involved to telecom operators. Calls made without this prior declaration will be treated as unwanted commercial communications, TRAI says.


Follow IMC 2026 with ET for live updates and insights.