Telcos have blocked over 20 billion spam calls with custom data, AI stacks—and they are just starting

Bharti Airtel, for instance, claims to have blocked over 20 billion spam calls in since September last year. Vodafone Idea said that it blocked over 600 million spam calls in the past three months. The two together account for 51% of India’s 1.2-billion mobile network users as of August, per data from the Telecom Regulatory Authority of India (Trai).

Though late to the effort, Reliance Jio, India's top phone services company by subscribers, has been rolling out spam detection on its network for over three months now and believes that it will move the needle.

“That said, detection and reduction of spam callers on our network will take time, because it involves verifiably tracking down spam callers and identifying the numbers. We’re already seeing that happening—and it’s only a matter of time before we get down to tracing and alerting spam calls and messages at least nine out of 10 times," said Sunil Dutt, president of devices, sales and distribution at Reliance Jio.

99.9% accuracy in spam detection

Its compatriots already claim high efficiency and accuracy figures. Rohit Misra, vice-president of digital operations at Vi, said that the company uses a hybrid approach for its anti-spam approach. “We have an in-house team that has developed a proprietary algorithm to track spam calls and messages but also use databases that are both internal as well as through public sources—such as the ‘Chakshu’ portal operated by the Centre. AI helps us monitor patterns and customer interactions with a phone number. Today, if we are alerting a user about a likely spam caller, 99.9% of the time this turns out to be accurate," he said.

A senior executive at Bharti Airtel said his company had invested millions to create a machine learning (ML)-based tech stack that's been trained across beats. “These include algorithms trained specifically to protect against banking and financial frauds, spammers marketing fraudulent products, and more. (The) stack uses a combination of 250 filters," he said, adding 99.9% of all spam calls a user may receive are detected. He requested anonymity.

The three telcos together account for 92% of India’s mobile phone users, with 205 million, 396 million and 484 million users on Vi, Airtel and Jio respectively. Their moves to double down on spam detection were notified in amended rules of Telecom Commercial Communications Customer Preference Regulations (TCCCPR), 2018. The rules urged telecom operators to proactively trace spam calls, accept complaints from users up to seven days after receiving a call, act on a complaint within five days, and more severe blacklisting rules for serial offenders.

Benefits beyond spam

Analysts believe that alongside regulation, the aggressive push from the operators to offer spam protection also comes from benefits for themselves.

Building and using a proprietary AI stack is a necessity for telecom operators, since only then will they be able to monetize vast troves of data with technology companies building large language models, said Siddhant Cally, telecom research analyst at Counterpoint Technology. The AI stacks built for proactive monitoring are "also necessary both from improving network operational efficiencies, as well as for ancillary marketing with AI firms such as Perplexity. The use of AI algorithms in spam detection and protection is a factor of all these use cases, and to this end, is a useful one for users," he said.

Cally added that while spam detection may not help telecom operators in user acquisition, “it is certainly an added incentive for user retention for any of the three top operators."

For now, while Jio remains in ramp-up mode, and Airtel and Vi claim high efficiencies, there are gaps still in the system. “One clear gap is the lack of compliance to use recommended business numbers among many enterprises. If they make the shift, compliance becomes easier for all parties. But, in the case of businesses using a series of other numbers, successful spam detection becomes a complicated affair," Vodafone Idea's Misra said.

The Airtel executive quoted earlier in this story added that the filters themselves are a complex mix in terms of operational accuracy of the spam detection process. “We’re using a combination of whether a person changes their SIM on their phone five times in a month, whether they make periodic outgoing calls only—without any incoming traffic, and more such factors," he said. Geolocating and locking a service to a SIM for automated background authentication is another tactic.