Telecom operators are being required to use Artificial Intelligence and Machine Learning (AI/ML) platforms to detect suspicious calling behaviour
The Telecom Regulatory Authority of India (TRAI) has once again tightened rules against spam calls, or unsolicited commercial communication (UCC). The regulator is no longer relying primarily on consumers to identify spam after they receive it. Telecom operators are being required to use Artificial Intelligence and Machine Learning (AI/ML) platforms to detect suspicious calling behaviour.
Under the new rules, telecom operators must identify Calling Line Identifications (CLIs) with a high probability of being used for UCC and share this intelligence across networks. If operators flag five or more CLIs associated with a sender within 10 days, they can initiate KYC re-verification, physical verification, bar outgoing services, and for repeat violations, disconnect telecom resources. The other important change is bringing application-to-person (A2P) calling explicitly into the regulatory net. Entities using auto dialers, robocalls or pre-recorded/artificial voice systems must declare their use and the CLIs they will employ. Undeclared A2P calls will be treated as UCC. More importantly, the regulator can now levy a termination charge on the originating access provider for A2P calls. At the same time, regulated commercial and authorised calls are exempt.
These measures build on a regulatory offensive that has gathered pace over the past two years. But spam calls remain a problem. TRAI’s complaint data show UCC complaints rising from 10.83 lakh in 2022 to 13.63 lakh in 2023, 19.39 lakh in 2024 and 31.09 lakh in 2025. In April-June, consumers filed 10.85 lakh UCC complaints, of which 5.53 lakh were actionable. But operators’ AI systems flagged 22.9 billion calls and 1.4 billion SMSes as suspected spam in the quarter. The numbers suggest that the success of regulations cannot be judged simply by the number of notices, disconnections, or complaints. TRAI needs a better measure: how many unwanted calls actually reach consumers before and after regulatory intervention. AI also introduces a new uncertainty. How accurately can an algorithm distinguish a spammer from a legitimate high-volume caller? TRAI should, therefore, publish the accuracy of its AI systems, including false positive rates and safeguards available to legitimate senders.
The problem is compounded by the fact that spam operators can adapt. Once a number is identified, callers can switch CLIs, distribute traffic across multiple numbers, automate calls or alter calling patterns. TRAI’s decision to focus on multiple CLIs associated with a sender rather than individual numbers recognises this problem. It has already begun moving enforcement upstream, instead of relying only on customer complaints after a spam call has been made. The next stage must be to translate AI-led detection into reliable prevention. TRAI should push for stronger caller ID authentication, real-time network-level blocking of high-confidence spam, better detection of SIM farms and number rotation, and continuous intelligence-sharing across operators.
Published on September 27, 2026
