Airtel has recently launched an AI-powered spam detection tool aimed at protecting users from unwanted calls and messages. While the tool promises enhanced security and a more streamlined communication experience, experts have expressed concerns over potential privacy implications.
The newly launched network-based AI solution is designed to automatically identify and detect spam calls and messages. The tool intends to help users recognise when a spammer is calling, allowing them to make informed decisions about whether to answer or reject the call. The telecom provider stated that the tool was developed internally.
Airtel's latest initiative comes in response to a number of unsuccessful efforts by telecom operators, tech companies, and regulatory authorities to tackle the issue of unwanted calls and SMS messages.
In light of the increasing incidences of spam calls, a nationwide survey by LocalCircles revealed that an alarming 77 percent of respondents receive three or more spam/scam calls each day. Additionally, a significant 96 percent of individuals who registered on the DND list reported that they still receive unwanted calls.
However, tech advocate Salman Waris pointed out that implementing such AI-based spam detection comes with its own set of challenges and considerations. He said, "Since AI-based spam detection relies heavily on analysing large volumes of data, which can sometimes include the content of messages, privacy concerns arise even if the actual content of calls is not monitored. Therefore, it is essential to safeguard user data and ensure privacy, particularly when handling sensitive information."
The telco's new AI feature is automatically activated for all prepaid and postpaid Airtel subscribers at no extra charge. While it is available on smartphones, the company is still exploring options to enable the AI spam detection capability for feature phones.
How will Airtel detect spam/scam calls?
Airtel is essentially counting on subscribers to avoid answering scam calls or clicking on dangerous links in SMS messages marked as 'Suspected Spam'. By encouraging users to steer clear of such tactics, the company aims to reduce the frequency of fraudulent communications.
To detect these scams, Airtel has implemented a two-tiered spam filtering system, with one filter operating at the network layer and the other at the IT systems layer. The company claims its proprietary AI algorithm analyses 250 different parameters related to calls in approximately 2 milliseconds.
Some of the indicators the AI will examine include call velocity, frequency of device changes, various call locations, unanswered call statistics, DND complaints, incoming and outgoing call patterns, call volume, SIM card change frequency, purchase locations, duration of calls, robocalling devices, the age of the network, the number of SIM cards registered under the same KYC, insights from cyber cells, and more.
To notify customers about harmful links received through SMS, the telecom company stated that its AI model will verify links against a centralised database of blacklisted URLs while monitoring frequent changes in the IMEI, a unique 15-digit number assigned to each device. According to its testing, Airtel reported that the AI model currently achieves a 97% accuracy rate in identifying spam calls and a 99.5% accuracy rate for spam SMS messages.
“In the next four to six weeks, we will see the AI model bridge the 3 per cent gap in accuracy further as it is a learning model,” Gopal Vittal, managing director and CEO of Bharti Airtel, said at a press briefing.
On comparisons with Truecaller, Vittal expressed confidence that Airtel’s AI spam filters offer a superior solution compared to it, as these crowdsourced apps depend on user feedback for spam detection and require internet connectivity to function. “Even if you’re on Truecaller, the calls are not being identified as Suspected Scam so you don’t even know. In our case, it will be detected in real-time,” he said.
What are the privacy red flags?
While the tool promises enhanced security and a more streamlined communication experience, experts have raised concerns about potential privacy implications and the unintended consequences of an AI solution.
In an interview with BOOM, technology advocate Utkarsh Kumar emphasised that although Airtel claims it does not monitor the content of calls, it does track call data records. This includes the collection of personal information, such as call frequency, duration, and the ratio of incoming to outgoing calls.
"However, Airtel has not provided strong safeguards or independent audits to ensure that this does not lead to profiling users into different categories. This poses a significant privacy risk, especially since Airtel is a private entity and not subject to the transparency standards required of public institutions," he stated.
Advocate Waris also stressed the importance of Airtel implementing robust data protection measures, such as encryption and access controls, to protect user privacy and comply with relevant data protection regulations.
He remarked, "Initiatives like the AI-based spam detection solution adopted by Airtel, which may impact user privacy, ideally should have been implemented based on user consent, as mandated by the Indian Digital Personal Data Protection Act, which requires a consent-based framework."
Waris further noted, "In the realm of cybersecurity, the battle against spam is continually evolving, with AI playing a crucial role. By utilising tools like false positive filters, telecom companies can enhance spam detection with the accuracy that AI provides."
Addressing the potential consequences of false positives, Kumar explained, "Consider the cases of lawyers, journalists, and salespersons who call numerous new contacts daily; their calls may be mistakenly flagged as spam by AI. The worst-case scenarios could involve emergency or critical information being misclassified, leading to unintended consequences."
Kumar suggested that such a model should have originated from TRAI rather than individual operators. This approach would ensure uniform implementation across all operators while also subjecting it to public scrutiny, he said.