Airtel Trying To Curb Spam/Scam Calls Using AI

Starting September 26, Airtel subscribers will notice a “Suspected Spam” banner appearing on their dialer screens whenever they receive a call from a number identified as spam or a scam by the telecom company's AI model. SMS messages containing potentially harmful links will also be flagged as suspected spam.

This new feature is automatically activated for both prepaid and postpaid Airtel users at no extra cost. While the focus is primarily on smartphones, Airtel is working on extending this AI-powered spam detection to feature phones.

Airtel's initiative comes after several attempts by telecom operators, tech firms, and regulatory bodies to combat unwanted calls and messages, including the establishment of a Do-Not-Disturb (DND) registry.

Gopal Vittal, the managing director and CEO of Bharti Airtel, stated at a press briefing in Delhi, “We have spent the last twelve months comprehensively addressing this issue. Today marks a milestone as we launch the country's first AI-powered spam-free network to protect our customers from the relentless barrage of intrusive communications.”

However, the question arises: what distinguishes Airtel’s approach from previous efforts? Is AI the key to finally alleviating the annoyance of unwanted calls and texts?

How Airtel Detects Spam and Scam Calls

Airtel's strategy depends on subscribers avoiding scam calls and not clicking on dangerous links in SMS messages flagged as suspected spam. As more users refrain from engaging with such tactics, Airtel aims to reduce the frequency of these fraudulent communications.

“The user will see ‘Suspected Spam,’ and the moment they do, they are fully protected,” Vittal explained.

To identify spam, every call and SMS will be screened through two filters: one at the network level and another at the IT systems level. Airtel asserts that its proprietary AI algorithm analyzes 250 different call-related parameters in approximately 2 milliseconds.

Indicators examined by the AI include call frequency, changes in device usage, various locations for calls, unanswered calls, DND complaints, call trends, call volumes, SIM card changes, and inputs from cyber cells, among others.

“By cross-referencing this information with known spam patterns, the system can accurately flag suspected spam calls and messages,” Airtel stated.

To identify malicious links in SMS messages, Airtel's AI model will check against a centralized database of blacklisted URLs, while also monitoring frequent IMEI changes—a unique 15-digit number assigned to each device. Addressing privacy concerns, Vittal clarified that the AI algorithm does not read SMS content but looks for spam indicators.

Based on its tests, Airtel reports a 97% accuracy rate for spam calls and 99.5% for spam SMS messages. “In the next four to six weeks, we expect the AI model to close the 3% accuracy gap as it learns,” Vittal added.

Differences from Apps Like Truecaller

Vittal expressed confidence that Airtel’s AI spam filters offer a superior solution compared to apps like Truecaller, which rely on user feedback for spam detection and require internet access. “Even if you’re using Truecaller, the calls aren’t identified as ‘Suspected Spam,’ so you’re unaware. Our system detects them in real-time,” he emphasized. He also noted that the Calling Name Presentation (CNAP) proposed by the Telecom Regulatory Authority of India (TRAI) may not be effective, as many subscribers might not have SIMs registered in their names.

Nonetheless, Vittal acknowledged certain limitations of the AI spam filters. Currently, the telco is not blocking spam or scam calls since about 3% may originate from legitimate businesses, such as Swiggy, which could mistakenly be marked as spam.

Recently, Vittal wrote to the leaders of rival telecom companies like Jio, Vi, and BSNL, suggesting they share data on business-to-business subscribers to better differentiate spam from legitimate calls.

As these AI spam filters are only applicable within the Indian network, they may not effectively address the rising number of international spam calls. Recent data indicates that over 85,000 scams targeting Indians this year have originated from Southeast Asian countries. Moreover, spam on Over-the-Top (OTT) communication platforms remains unaddressed. “We can’t tackle spam on OTT apps. They need to be brought under regulation,” Vittal remarked.

Can AI Effectively Combat Spam and Scam Communications?

In June, TRAI directed all telecom service providers to implement AI/ML spam detection systems to reduce unsolicited communication from unregistered telemarketers (UTMs). The directive emphasized the need for systems capable of evolving to counter new signatures and tactics used by UTMs, following previous unsuccessful measures.

Experts have previously warned that AI spam detection systems may not suffice without stringent penalties for repeat offenders. TRAI has also proposed a discussion paper suggesting differential tariffs for subscribers making or sending over 50 calls or messages daily.

When asked if spammers identified through Airtel’s AI model would be reported to regulators, Vittal responded, “We are open to sharing data regarding legitimate scams or spam. This solution aims to prevent scams by ensuring that the calls don't occur, as every scam begins with a call.”