AI evolution has elevated our predictive capabilities against fraud: Truecaller India MD 

Sweden-based caller identification app, Truecaller, has just introduced its artificial intelligence (AI)-driven assistant in India. To get an inside look into how AI is seamlessly integrated into the platform and how it's making the app even more powerful, TechCircle spoke to Rishit Jhunjhunwala, Chief Product Officer & Managing Director of Truecaller India. In this discussion, he shared how Truecaller employs AI to enhance communication security, tackle spam, and offer other assistance features. Edited excerpts:  
 
How is AI integrated into your tech services and enhancing Truecaller's capabilities? Can you provide details on how Truecaller is utilising AI?

AI's widespread adoption and discussion stem from recent advancements in data utilisation. Over time, it has found varied applications. A prominent use is in enhancing communication security for calls and SMS. Battling spammers and fraudsters requires constant innovation, wherein predictive AI models play a crucial role even before user input. These models categorise and identify potential scams.   
SMS spam and fraud detection also leverages the AI model, learned through user feedback, which resides within the Truecaller app, discerning spam messages on the user's device, without external data transfer. This offline model identifies not only spam but also categorises messages (e.g., bank SMS, flight info) locally, boosting user privacy. AI's evolution has elevated our predictive capabilities against fraud. AI's benefits extend to customer support too among other areas.
 
Why did your team decide to focus on transcription as a feature for your AI chatbot, considering the presence of other AI assistants and systems in the market? 

We consistently seek to expand Truecaller's capabilities beyond caller identification and scam detection. An adjoining concept emerged — the potential to ascertain incoming calls without direct engagement, understanding the caller's intent, and even facilitating brief interactions. We had acquired Call Hero — the AI assistant tool that was rebranded as Truecaller Assistant after the acquisition by Truecaller — to integrate its technology into the Truecaller platform. Initially, we unveiled the assistance feature in the United States, catering predominantly to the English-speaking demographic on both Android and iOS platforms. Gradually, we expanded its availability to other English-speaking markets, all the while adapting and customising the offering for the unique needs of each region. Simultaneously, we worked on localising the product for the Indian market.

Indian market lacks even basic features like voicemail, which is standard in Western countries. Introducing such features in India raises challenges due to the unfamiliarity of the concept among users. Nevertheless, we believe that the Indian populace has grown accustomed to leveraging technology for improving their lives. Currently, the assistance product supports English, Hindi, and Hinglish. As the service needs to comprehend various languages, we intend to expand its language support significantly. Indian Truecaller user base is now ready to anticipate advanced capabilities like assistance services, call recording, and improved spam detection.   
  
How do you ensure data privacy and security for services like caller identification and spam blocking, especially considering your data leak incident in 2019?  

Recently, we introduced the SMS Fraud Protection service. Our primary goal was to ascertain the feasibility of constructing this feature without necessitating any data transmission from the device, striving for complete offline functionality. If minimal data exchange with our servers becomes necessary, we adhere to a set of stringent policies. First among them is a user opt-in requirement for any data retrieval; we do not acquire data without explicit consent. Subsequently, we rigorously limit the transmitted data to the bare essentials. For instance, if it's plausible to transmit only metadata, we confine the exchange to that without involving core data.

Once data reaches our servers, we implement measures to further minimise it by purging it after a stipulated time period. Although regulations mandate the retention of certain data, we voluntarily discard it beyond this period. Our approach is designed to maintain user control at all times. If a user opts not to appear on Truecaller, they can exercise this choice.

Our Application Programming Interfaces (APIs) are not exposed in any form. Data access is limited strictly on a need-to-know basis and to a select few within the organisation. Moreover, back in 2017-2018, we took the initiative to localise Indian subject data within Indian borders, underscoring our commitment to data protection.

Unlike certain entities, we eschew practices like data monetisation, even when dealing with aggregated data. Our financial stability, demonstrated by profitability since 2018 and our listing on the Swedish stock exchange, corroborates our ability to maintain independence without resorting to such practices.
  
What are new initiatives at work?

A multitude of features consistently undergo beta testing within our development pipeline. We maintain a keen focus on AI integration and exploration. This commitment extends beyond individual consumers to encompass our business-facing offerings.