AI Automation in Telecom_ Ensuring Accountability and Public Trust India AI Impact Summit 2026
20 Feb 2026 10:00h - 11:00h
AI Automation in Telecom_ Ensuring Accountability and Public Trust India AI Impact Summit 2026
Summary
The panel examined how AI-driven operations can be leveraged to build and preserve customer trust in telecom services, noting that AI already shapes outcomes such as outage management, grievance handling and spam-fraud prevention, but must balance efficiency with false-positive reduction and privacy compliance [11-15][16-18]. Speakers emphasized the necessity of a “human-in-the-loop” to keep AI decisions from running unchecked [19-20].
Julian Gorman described the “scam economy” outpacing regulation and presented GSMA’s Cross-Sector Any-Scam Task Force, which has collected over 40 operator case studies and is piloting data-sharing proofs of concept across Asia-Pacific [26-38]. He warned that service-based rules can stifle future innovation and argued that regulation should focus on outcomes while fostering industry-wide collaboration, especially as India assumes a global telecom leadership role [39-53].
Dr Rajkumar Upadhyay showcased CDOT’s AI suite, including “Fraud Pro” that de-duplicates SIM registrations and has disconnected 70 lakh fraudulent connections, a digital intelligence platform used by banks for financial-risk scoring, and the crowdsourced “Chakshu”/“Sanchar Sati” app with over 18 million downloads that empowers users to report and block unwanted calls [65-71][73-102]. He also described an AI-federated disaster-management system that integrates alerts from IMD, CWC and other agencies, uses cell-broadcast for geo-targeted warnings, and has reduced cyclone-related deaths in Odisha to zero, a model now being promoted to the UN for global rollout [250-266][270-276].
Mathan Babu Kasilingam explained that his telecom operator follows privacy-by-design principles, holding ISO 27701 certification, and views AI adoption through pillars of responsibility, reliability, trust and privacy [113-119]. He noted early “quick-win” AI projects in fraud detection and network self-healing, but identified siloed data repositories and high infrastructure costs (80-90 % of AI spend) as major obstacles, prompting a shift toward a unified data platform and enterprise-wide LLMs [124-138][149-170][225-230]. Centralising data also simplifies compliance with India’s DPDP rules and enables scalable AI model refinement [184-187][219].
Syed Tausif Abbas introduced a voluntary AI-incident reporting schema with taxonomy covering network components, severity and cause, which can help operators analyse failures and inform regulators [193-201]. Both Abbas and Kasilingam agreed that such standardized reporting can streamline model improvement and support cross-border data sharing, a point echoed by Gorman who called for regulated sandboxes, open-gateway APIs and four pillars-network security, ecosystem exposure, customer services and digital skills-to combat scams collaboratively [201-207][285-306].
The discussion concluded with consensus that coordinated global effort, especially data sharing across borders and adherence to emerging standards, is essential for responsible AI deployment in telecoms [321-326][327].
Keypoints
Major discussion points
– AI must be harnessed to protect customers and preserve trust.
Dr. Tangirala emphasized that AI-driven decisions affect users (outage management, grievance handling) and that clear, proactive communication is essential. He highlighted the tension between aggressive fraud-spam reduction and the need to avoid false positives while respecting privacy and regulations, and called for a “human-in-the-loop” safeguard [13-18][19-21].
– Industry-wide, cross-sector collaboration is critical to combat scams.
Julian Gorman described the “scam economy” as faster than regulation and outlined GSMA’s Cross-Sector Any Scam Task Force, which has gathered >40 operator case studies and is developing data-sharing proofs of concept. He stressed that regulation should focus on outcomes and that global cooperation-especially India’s emerging leadership-is needed to keep innovation alive while fighting fraud [26-35][36-44][45-53].
– AI-powered solutions from CDOT illustrate concrete use-cases for fraud, identity-deduplication, financial risk, and disaster management.
Dr. Rajkumar Upadhyay presented tools such as “Fraud Pro” (detecting duplicate SIM registrations), a digital intelligence platform for banking risk scores, the crowdsourced “Chakshu” app, and an AI-driven early-warning system that fuses meteorological data and cell-broadcast alerts to save lives [65-73][74-84][85-102][241-276].
– Service providers are grappling with AI adoption, data silos, infrastructure costs, and privacy-by-design.
Mathan Babu Kasilingam explained the provider’s journey: certification under ISO 27701, the shift from isolated data lakes to a unified AI platform, the challenge of massive GPU/compute spend (≈80-90 % of AI cost), and the need to balance “quick-win” pilots with a consolidated, secure architecture that can support LLMs and self-healing networks [111-119][120-138][148-166][184-188][225-236].
– A voluntary AI-incident-reporting standard (TEC) was introduced to enable systematic learning from AI failures.
Syed Tausif Abbas outlined the schema (30 fields covering incident type, severity, affected subsystem, etc.) and argued that, although not mandatory, the standard would give operators a common data set for root-cause analysis and help regulators shape AI policy [193-199][200-203][204-209][210-218].
Overall purpose / goal
The panel aimed to explore how AI can be responsibly deployed across telecom operations to build and sustain customer trust, by sharing best-practice use-cases, highlighting the need for collaborative standards and cross-border data sharing, and discussing practical challenges (privacy, cost, governance) that operators, regulators, and standards bodies must jointly address.
Tone of the discussion
– The session opened with a formal, courteous tone (introductions, opening remarks).
– As speakers presented concrete AI applications and the urgency of fraud/scam mitigation, the tone became focused and problem-oriented, yet remained constructive, emphasizing solutions and collaboration.
– When the voluntary standard and cost-optimization topics were raised, the tone shifted to analytical and forward-looking, acknowledging hurdles while expressing optimism about unified platforms and global cooperation.
– The closing remarks returned to a appreciative and collegial tone, thanking participants and reinforcing the collaborative spirit of the forum.
Overall, the conversation maintained a professional and solution-driven atmosphere, moving from introductory formality through detailed technical discussion to a concluding note of mutual respect and shared commitment.
Speakers
– Dr. M P Tangirala
– Area of expertise: AI in telecom, customer trust, responsible AI
– Role / Title: Chair of the panel / Session moderator (as introduced to begin the session)
– Anil Kumar Jha
– Area of expertise: Telecom regulation, policy advisory
– Role / Title: Principal Advisor, Telecom Regulatory Authority of India (TRAI) [S2]
– Mr. Julian Gorman
– Area of expertise: Telecom industry collaboration, anti-scam initiatives, AI-driven security
– Role / Title: Representative, GSMA (Asia-Pacific) – expert in telecom collaboration and scam mitigation [S4]
– Syed Tausif Abbas
– Area of expertise: Telecom standards, AI incident reporting standards, policy formulation
– Role / Title: Senior Deputy Director General (DDG) and Head, Telecom Engineering Centre (TEC); also holding additional charge as CMDTCIL
– Dr. Rajkumar Upadhyay
– Area of expertise: Telecom AI applications, fraud detection, disaster-management systems, quantum communications
– Role / Title: CEO, Centre for Development of Telematics (CDOT) [S8][S9]
– Mathan Babu Kasilingam
– Area of expertise: AI adoption in telecom service providers, privacy-by-design, AI infrastructure, LLM integration
– Role / Title: Senior executive representing a telecom service provider (TSP) – speaker on AI adoption and privacy standards [S11]
– Moderator
– Area of expertise: Technology security, data privacy, cyber-security governance
– Role / Title: Technology Security and Data Privacy Officer, Vodafone India Limited (over 20 years experience) [S12]
Additional speakers:
– None. All participants in the discussion are covered by the speakers list above.
The session opened with a formal introduction by the moderator, who identified the Technology Security and Data Privacy Officer of Vodafone India and senior DDG S.T. Abbas as panelists and invited the audience to focus on “balancing information, innovation with privacy and trust” before handing over to Dr M P Tangirala to chair the discussion [1-8].
Dr Tangirala set the tone by stressing that AI-driven decisions-whether in outage management, service continuity or grievance handling-directly affect customers and therefore require clear, proactive communication [13-21]. He warned that while AI can dramatically improve fraud-spam reduction, it must be deployed so as to minimise false-positives and fully respect privacy and regulatory constraints [16-18]. A “human-in-the-loop” safeguard was presented as essential to prevent autonomous systems from making unchecked decisions [19-21], and he announced that the panel would hear from experts representing service providers, R&D and the DOT standard-setting body [22-25].
Julian Gorman described the “scam economy” as a threat that moves faster than regulation, noting that scammers are not bound by geography, law or funding limits [26-30]. To counter this, GSMA created the Cross-Sector Any-Scam Task Force, a coalition of more than 39 organisations from 17 countries-including Meta, Google, TikTok and AWS-aimed at identifying and prioritising joint anti-scam initiatives [31-35]. He reported that, within a few months, over 40 operator case studies from the Asia-Pacific region had been collected, demonstrating that operators can develop and implement successful scam-mitigation strategies without waiting for regulation [36-38]. Gorman argued that service-based rules risk stifling future innovation and that regulation should focus on outcomes while fostering industry-wide collaboration, especially as India rises to a global telecom leadership role [39-53].
Dr Rajkumar Upadhyay presented CDOT’s AI portfolio. He began with “Fraud Pro”, a system that groups images and demographic data to detect duplicate SIM registrations-an approach that has already disconnected 7 million fraudulent connections [65-71][73-84]. He also described a digital intelligence platform used by banks to assign a risk score to transaction recipients, enabling financial-risk indicators that block high-risk transfers [81-85]. The crowdsourced “Chakshu”/“Sanchar Sati” app, downloaded by more than 18 million users and generating 25 crore website hits, empowers customers to report unwanted calls and automatically disconnects fraudulent numbers, with 7 million connections removed through user-initiated verification [85-102]. Upadhyay highlighted that the AI platform was used to locate dead bodies after the Balasore train accident, demonstrating AI’s utility beyond telecom services [80-85]. In the domain of public safety, CDOT has built an AI-federated disaster-management platform that aggregates alerts from IMD, CWC and other agencies, uses AI to generate geo-targeted cell-broadcast messages (a technology that sends alerts to all devices in a geographic area), and has reduced cyclone-related deaths in Odisha to zero-a model now being promoted to the UN for global early-warning deployment by 2027 [241-266][270-276].
Mathan Babu Kasilingam outlined his operator’s AI governance framework. The company is certified under ISO 27701 (Personal Information Management System) and follows a “privacy-by-design” approach, positioning privacy as a core pillar of trust [113-119]. He traced AI adoption from the consumerisation of assistants such as Siri and Alexa to enterprise quick-wins, where AI is first applied to a single function (e.g., fraud detection) to demonstrate value [120-138]. He identified two major obstacles: fragmented data silos created by separate AI projects [148-166] and the high cost of infrastructure-80-90 % of AI spend goes to GPUs, storage and compute [225-230]. To address these, the operator is consolidating data into a single AI lake, exposing it through enterprise-wide APIs, and developing purpose-built large language models (LLMs) that can serve multiple business functions while simplifying DPDP (Data Protection and Data Privacy) compliance [184-188][170-179][219-224]. Kasilingam noted that his organisation already records incidents within its ITIL-based processes and sees the TEC schema as complementary to existing practices [193-197].
Syed Tausif Abbas introduced a voluntary AI-incident-reporting schema devised by the TEC standard-setting body. The schema comprises 30 fields covering incident type, severity, affected subsystem, cause and impact (physical, environmental, psychological), with submitter details masked for privacy [193-201][202-207][208-218]. Although the standard is not mandatory, Abbas argued that a common taxonomy will enable operators to analyse failures, refine models and provide regulators with consistent data to shape AI policy [193-197][201-207]. He likened the initiative to the early computer-emergency-response teams, suggesting that a similar mechanism is now needed for AI [195-197].
The subsequent Q&A reinforced several themes. Kasilingam emphasized the value of quick-win projects and discussed the cost pressures of AI infrastructure [225-236]. Upadhyay expanded on the disaster-management system’s scalability and its potential for international adoption [241-266]. Gorman reiterated the need for privacy-enhanced cross-industry data sharing via open-gateway APIs and regulatory sandboxes [285-295]. In response to Mr Jha’s query, he proposed two global steps-cross-border secure data sharing and collective action-and two India-specific steps-domestic anti-scam measures and knowledge export [321-326].
The moderator concluded by thanking the panelists for their “vibrant discussion on responsible AI, standards, the repository and various government apps for enhancing consumer experience,” and announced the presentation of mementos and a group photograph, underscoring the collaborative spirit of the session [328-335].
Key take-aways
– Human-in-the-loop controls and proactive customer communication are essential for trustworthy AI-driven telecom services [13-21].
– Privacy-by-design, demonstrated by ISO 27701 certification, builds customer confidence [113-119].
– A voluntary AI-incident-reporting schema with a 30-field taxonomy can increase transparency and aid regulators [193-201][202-207].
– Cross-sector collaboration-exemplified by GSMA’s Any-Scam Task Force and privacy-enhanced data-sharing sandboxes-is critical to combat scams [31-35][285-295].
– AI-based fraud tools such as “Fraud Pro”, “Sanchar Sati”, and millisecond-level call-blocking have demonstrably reduced fraudulent connections and scam calls [65-71][85-102][307-309].
– The AI-federated disaster-management system that fuses meteorological data and delivers geo-targeted cell-broadcast alerts has achieved zero-casualty outcomes in pilot regions [241-266][270-276].
Action items
– GSMA will continue expanding the Cross-Sector Any-Scam Task Force and its Southeast-Asia proof of concept [31-33].
– CDOT will promote its fraud-prevention suite, disaster-management platform, and the dead-body detection capability for international adoption [65-71][241-266][80-85].
– Kasilingam’s organisation will merge fragmented data repositories into a single AI infrastructure and expose services via enterprise APIs [152-166].
– Abbas will circulate the voluntary incident-reporting schema to encourage uptake [193-201].
– Gorman recommended establishing privacy-enhanced data-sharing sandboxes and regulatory support mechanisms [285-295].
Unresolved issues
– Detailed regulatory frameworks that enable privacy-preserving cross-industry data sharing [285-295].
– Strategies to address the shortage of skilled AI talent within telecoms [225-229].
– Methods to balance aggressive call-blocking with the guarantee of emergency call availability [307-309].
Technology Security and Data Privacy Officer at Vodafone India Limited, with over 20 years of experience in cyber security domain and governance structure. Rounding off the panel, we welcome Mr. S .T. Abbas, Senior DDG and Head TEC, also holding additional charge CMDTCIL, with over 35 years of experience in telecom standards, certifications, spectrum management and network regulation. I would request all the panelists to please come forward for a quick photograph Thank you, sirs. Please take your seats. Let’s engage deeply on how to balance information. Innovation with privacy and trust. I now hand over to Dr. Tangirala Ji to begin the session. Thank you.
Chairman, member, Mr. Mitter, distinguished delegates, my fellow panelists, I welcome everyone to this second session. The clock is already ticking, so I will be brief in my opening remarks because I come between the audience and the distinguished panelists, which I don’t intend to do. The session title is Building Customer Trust Through AI -Driven Operations. The importance of trust was highlighted, among others, by Mr. Shantigram Jagannath as well, when he was speaking about AI through telecom networks and the at -scale problems that we could try and solve. Thank you. Now, while customers may not interact with AI models directly, they are affected by the outcomes of the decisions. And therefore, you know, whether it’s outage management, service continuity, grievance handling, you know, while efficiencies may improve, the responsibility for decision integrity ultimately remains with the telecom service providers.
And clear and proactive communication with the customers would become very important. And that is where, you know, there are impactful applications of AI in telecoms, in spam and fraud prevention, which a person had mentioned in his opening remarks about how 2 .1 million numbers were disconnected using AI -based tracking. But the challenge is also that we need to reduce this spam. while minimizing false positives, avoiding customer inconvenience, and fully respecting privacy and regulatory requirements. So that is always a big concern. Then, of course, this whole issue of the human in the loop or human in the mix. We need this automation to have an element of human control that is so that the system does not run away with its own decisions.
So we have, for all these issues and more, we have eminent speakers here, both from the service providers, from the R &D, and as well as from the standard -setting body of DOT. I will request each of them to give their thoughts, and then maybe a few of you… Both of them have presentations to make. I’ll request them to keep it to about five minutes or so, so that we have time for further discussion. Thank you.
And the reason for it is in the scam economy, regulation cannot move as fast as scammers. Scammers are not bound by geography. They’re not bound by laws. They’re very technically capable and they’re very well funded. They have all the things that mobile operators would like to have. I think it’s important to understand that we have to focus on stimulating innovation. At GSMA, about 12 months ago, we formed a coalition called Cross -Sector Any Scam Task Force. It involves more than 39 organisations from 17 countries, including the social media platforms, so Meta, Google, TikTok, AWS. And the aim was to drive or identify and prioritise initiatives and activities that we could do as an industry to help combat. Now, one of those activities was let’s gather what the industry is doing.
Now, in just the last couple of months, across Asia Pacific, we’ve gathered case studies of more than 40 instances where operators without regulation have developed, implemented, and used successfully some sort of strategy or service to combat scam. And I think that’s an indication, along with GSMA’s globally working with people like Virginia Tech and with our foundry, with our proof of concept around data sharing, is the industry is focused on this. And the danger, of course, of implementing service -based rules is they restrict innovation in the future. And so we really need to focus on outcomes when it comes to regulation. And I think we all universally subscribe to the fact we need to combat scam. We need to work together.
And it’s not just the people in this room. We need to collaborate and work across the ecosystem. to make that possible. I think those principles actually also apply in the broader sort of sense of the term is how do we grow 5G, how do we make 5G meaningful to the whole economy, to all users. It’s about stimulating that ecosystem and making sure that they are using 5G and 4G and mobile broadband into meaningful solutions for the population. And the important thing also for India is India is rising not just economically but also in its position in the telecom world and the GSMA sort of global ecosystem is India is a real telecom superpower and it’s on the rise.
And that means actually it cannot just be worried about its domestic situation. actually it has to embrace that statesman role to be a global leader. And so actually considering cross -border, how does India play its role in a global ecosystem are critical to actually the sustainability and growth of the global ecosystem of which India’s vision is dependent on. It cannot exist alone. And I think it’s important that when we focus on innovation and solving things like scam, it is as part of a global community. It’s not just a national community. And so the actions we take, the innovations we look to stimulate have to be part of that global solution. Thank you.
That was thought -provoking, some of the things that you said about collaborative innovation or innovation through collaboration. We will come to that in a bit when we go for the questions. So, with that now may I request Dr. Rajkumar Upadhyay, CEO of CDOT for his presentation and opening remarks.
Respected Chairman, Mr. Lahoti, Mr. Mittal, Mr. Tanglura fellow panelists, industry leaders, policy makers experts, ladies and gentlemen thank you for inviting me here I think in the previous session there was talk about how do you optimize your network how do you self heal your network how do you make correction in the network so I’m not going to talk about that even though we also as India we have developed our own 4G and 5G we used because we were the late comers so we used quite a bit of AI in terms of predicting the faults because a lot of logs are generated by various systems so I’m not going to talk about that I’m going to talk about where is the PPT?
where is the PPT? So I’m going to talk about some use cases which we have developed during last few years. We are CDOT. We were established in 1984, and we had the legacy of developing the rural telecommunication. We work in primarily three to four areas, the mobile wireless, cyber security, information security that is done by quantum. Quantum AI is a horizontal thing and advanced telecom applications. But I will focus on these are our product line, and all of these products actually use AI because AI is so pervasive. Without AI, you cannot function. So all these product lines, whether it is mobile, whether it is cyber security, information security, disaster management applications, are using AI in a big way.
So one of the key application, key product what we have developed is Fraud Pro. What it does, it actually detects the fraudulent connections in the system. I think you may be aware with the cases of Jamtara, Mewat and all these sim factories running. And these sim factories were destroyed by this particular software. What it does, it groups all the images of the same person because if you go and buy 500 sims using same Aadhaar card or same driving license, this is what was happening. So it detects that and it not only matches the images, it also matches the demographics, name, father, name. And sees that whether names and photos are same but names are different. So using, I will come to the number.
I think some number was described in the beginning that how many connections were disconnected using this software. So this is deduplication and finding the, and in fact we developed for telecom, it is being used now, going to be used in driving license, passports, income tax, deduplication, Manrega deduplications. The second one, I think this, it mentions the AI. AI analysis, 86%. 7 crore mobile numbers. and it was very well used even to, you know, find out dead bodies in the Balasore train accident. The first use case of this particular platform was to identify the dead bodies. The second one is financial risk indicator. I think you would have seen in newspapers RBI has mandated the banks to use financial risk indicator.
What it does, if A is transferring money to B, so B, the credentials of B are checked with the platform which we have developed, which we call digital intelligence platform. And the platform returns a figure that this is risky number, is medium risk or low risk. If it is a high risk number, the bank will not happen, let that transition happen. And it has saved a lot of fraud cases currently. And all the banks actually are using this FRI, which is able to tell that the B number where the money is going is a, you know, dangerous number or a well -identified fraudulent number, the money is stopped. The next is the Chakshu. Chakshu is again crowdsourcing platform.
wherein if you get a fraudulent call or a promotional call or any kind of call or fake KYC or faking as police, you can report and using crowdsourcing, we are able to disconnect the number and we are able to take. And this is again using our Sanchar Sati app. Just to bring to the notice of the audience, rarely a government app has a download of 18 million plus. This is 18 million plus downloads are there. The hits in the website of Sanchar Sati are 25 crore. Very rarely you see that. So this shows the popularity of how customers are protected using the AI -based platform. This is again AI -based platform. This Tapcoff and CIR here, I don’t know how many of you have used.
I would request those who have not used, please use it. In Sanchar Sati, you go, you give your mobile number, it will tell you all the connections under your name. using fuzzy logic, fuzzy AI and fuzzy logic. It doesn’t ask you any other detail. And that number also we ask because we want to verify that it is you and the OTP sent. Otherwise, no other details are asked. Just by the detail, we are able to find out how many numbers are there. And just to bring to your notice, 70 lakh connections have been disconnected using this. People have themselves disconnected their numbers because it also tells you, this is not my number, disconnect it. This was a big problem for us.
When we blocked the SIMs in the country, these guys went outside. And they started pumping calls using Indian numbers. It’s spoof call. This technology is available. I can get call from my own number in using this. We were getting 15 million calls per day, 15 million. And now you see, this was a very complex system because when the call hits at the gateway, the system has to decide within millisecond that this call to be let it go through or block it. it has to be decision has to be millisecond and it has to be zero error because no actual call should be blocked and today we are able to do our because the rigorous testing happened with all the operators today we are able to we have totally neutralize this of course they have found another way they are they have taken sims in places like Cambodia Indonesia Myanmar from there they’re calling so again the AI based system is alerting that these are the numbers of that country and we are alerting the governments of that country AI based security solution because cyber is another major area for all of us and somebody was mentioning the AI will do cyber attacks and it’s true we see in our system AI attacking the systems earlier it was human and now it is fully AI so you have to use AI to counter it so the cyber security solution today what we provide is fully AI based so that it can coordinate between various particular solutions disaster management we have used AI you may be aware that India has deployed ITU CAP based disaster management system as well as 3GPP cell broadcast based disaster management system which is implemented across India we use AI here because what happens like IMD is giving me a warning on rain CWC is giving me a warning on flood weather report is coming so we federate all these inputs and using AI you have less than 2 minutes so this I won’t go through NMS of course we use AI to see that how the when the network is likely to be down and this is actually implemented in Bharatnet 1 and 2 it tells you that this is likely to fail this router is misbehaving or this node is misbehaving so that was my last slide so in nutshell what I am saying that the a lot of AI applications are needed for the customer side to protect the customers and we have made India has made a good progress in terms of reducing the frauds reducing the fraudulent connections therefore safeguarding the customers and we will be very happy to take these technologies to any part of the world given it is implemented at India scale thank you so much
thank you Dr. Rupajay that was very interesting the flavor of the kind of R &D that has been done and the apps that have been developed we now move to someone on the panel Mr. Matan Babu Kashi Lingam who is representing the service providers and you carry a lot of burden of customer expectations on your shoulders so do tell us about your thoughts on the topic of today’s thank you thank you
So a few things that we have done as a service provider, majority of the topics are touched upon from fraud and cyber security. I’m trying to say the role of AI in terms of establishing trust, our entire whole ecosystem, which is telecom ecosystem, relies primarily on the customer trust. So to ensure that we have given trust journey for our customers and in the means of adopting AI, there are various core secure pillars that we have followed through. Any AI adoption should have the reasonability, reliability, trust, privacy should come to deliver that. So as a TSP, when we have embarked on the journey of AI, that’s the first and foremost core element that we have taken into consideration.
We are one of the TSPs in the country who have taken the journey of privacy since past five plus years now. We are completely certified on PIMS ISO 27701. We are the only TSP in the country who have governed privacy by design and have certified ourselves against that as well. So that is to only ensure that the trust is given back to the customer. Now I will come back on the journey of AI adoption. First thing that has happened is consumerization of AI. AI has been part and parcel of our life since all of us learned about Siri, Alexa. Day to day home we have been living with AI for many, many years now. So the consumerization of AI happened many, many years back.
What happened in enterprises, there came the pressure of adoption of AI in enterprise. In that, the first and foremost thing that we did is we took as applied in consumer. Let us try and adopt it in AI. Enterprises as well. well. Obviously, it has its own benefit. The benefit being it shares quick win, right? So you get to see a first yearly win that you are able to see by deploying AI in your setup. So how enterprises embarked on that journey is you pick and choose one department, one function, one key problem that you are faced with, deploy AI in it, and you see results. So we saw all of these examples. Fraud. It’s a serious problem for the entire country as a whole.
What can we do? Can we leverage AI? And AI is capable of giving me million eyes and million hands in name of a single human operating that, right? So the power of AI came to aid. We are able to today identify fraud. Sir also briefly touched upon cyber security. So as national critical infrastructures as what we are as TSPs, today we are pressed with serious amount of attacks. So India apparently in the past one year has hosted as many mega significant events. whether it is G20, whether it is Mahakum, then there is situational geopolitical tensions that we went by and now we are hosting AI summit. So national critical infrastructures like TSPs are also faced with increased volume of cyber attacks increased volume if I tell would be not 10 times it would be in as many count that I could multiply with that is a quantum of increased cyber attacks that we are seeing now in the cyber field we are also limited with the number of professionals we have.
So the power of AI not just for the attackers as defenders as well we have started leveraging how can we leverage the power of AI to combat them. So we have, so those are quick wins right? Network operations with the advent of 5G we wanted self operating self healing networks. So in various smaller smaller areas where AI can be embarked that we can realize a very very quick business value, enterprises started adopting That’s the first part that we wanted the quick win, we saw the quick win. The challenge that came with that is we started seeing them in piecemeal approach. The data that we were working upon was almost similar. You gather this intelligence information from the same network elements and nodes.
But you started to look upon through different lenses. All that I need to do is to look through different lens. But I started creating individual siloed repository of data. So if you look at corporates today that have embarked AI, you will see as many isolated silos of data created for them. Because each of them want their lens. And for every lens, they didn’t see the data through the lens. They created a total isolation of the data. Second thing that happened is mammoth amount of infrastructure. Anybody who touches AI today talks about GPUs, humongous. Power that is required to run, etc., etc. That again at enterprise. Thank you. it is siloed data, siloed infrastructure that has been taken into account.
So the journey that we are today in is we had the quick wins, we have taken the first few steps, but we are re -looking at from a different standpoint as we see currently. So we have stepped back. Is there data deduplication that can be done today? In lieu of 20, 30 silos that I have created, do I want to create one single repository of this data? Thereby the secure element also becomes easier. If in silo I have to secure everywhere, bring them in one area, I have the ability to secure them well. Can I leverage a common platform infrastructure, which is the AI infrastructure that is required to put the data and then do these work?
We are doing that. So you can still leverage a comprehensive LLMs, individual businesses in variety of functions, I have taken their own purpose -built LLMs, right? Because… You will have a HR function. The provider for HR is a specific, say SAP, would be primarily driven for HR. And surrounding systems which are talking AI would have built on top of it. There will be a self -healing network. The network provider builds an AI -driven system. So there we are now stepping back to see can we build a comprehensive central LLM, which will still deliver the purpose that are required for looking at. So at V, the premise is core infrastructure, put data in comprehensive, expose them through interconnected enterprise API architecture, thereby businesses and users do not have to talk to the data directly.
They talk through the enterprise model, touch the AI infrastructure, and go and reach back the data for various reasons. So it could be to service my service provider. It could be to service my customers. It could be my customer support. Thank you. bridging them. That’s a platform journey that we are doing that. So this consolidation, like I told, privacy is by design. We are able to do the DPDP compliance inclusive, which is minimize the data. Data in one area, we are able to minimize them as appropriate. That’s what I wanted to share with. Thank you.
Thank you. Fascinating. Now we come to Mr. Abbas, who is the senior DDG from TEC standard setting. He’s promised that he will make a different presentation. So over to you, Mr. Abbas.
the name of application what are the technology used what is the purpose like that then what was the impact or harm information with the what was the incident like physical harm environmental property psychological so these things also forms part of the 30 key fields in which the input is to be given for the schema and then some of the information which is to be masked later on so those related to the name of the submitter email and other things related to submitter information which will be redacted later on similarly the taxonomy as earlier I told that it will classify the incident into different categories depending upon the incident type whether it is a subcategory as network description service quality outage or it is a security beach or AI mismanagement or then affected system whether core is affected whether the radio access network is affected whether the edge is affected or IOT components or physical so which part of the network is affected or any application which is related to user is affected and then what is the incident severity whether it is critical high moderate or low so that also will be recorded and cause of failure if it is known to the user otherwise the deployer or the service provider has to enter this what was the cause of the failure so basically this database will give input to the service provider also that they themselves can examine it they can analyze it and then realign their AI related application so that these incidents don’t recur in future so it’s a gradual auto development of their own AI system which will be then error free and gives the best results output so for this is only this standard has been made but it is not going to give any mitigation mechanism or something.
This is to be decided by the deployer who has deployed those AI application and it is not mandatory. Just as a beginning, when the new computer system came, initially there was not much thing but when the incident started then computer emergency response team was proposed and it started working on collecting the data related to the computer incident. So similarly since the AI has already begun, so we should have this mechanism in place so that we can have the AI incident reporting database also available. Thank you so much.
Thank you so much Mr. Abbas. Presenting arguably and congratulations arguably world’s first time that such a standard has been put out. So since we are fresh off with you, I will start with a question for you about what you have just now presented. you said it is not mandatory it is voluntary of course we will see where that journey goes as you said about certain coming in after the computers but can you tell us a little bit more about what value it offers to the telecom service providers if they voluntarily adopt this standard
so telecom service providers they have already started using the AI application in their network optimization network and services to the users orchestration of resources so many things already the AI application has started so if any incident which gives an unintended outcome if it is recorded and reported then it will be in the best interest of the service provider that those incidents are analyzed and then rectified for so that it is not occurring in the future so in this way it is a can be best utilized by the service provider and since the structure of the schema and taxonomy both are given so it will be a same structure compilation of data which every service provider is doing so that will give benefit to the regulator and policy makers to how to go about the AI policy because of those input which we get from those incidents.
So therefore, Mr. Kashi Lingam, would you think it offers a voluntary adoption of this standard offers any benefit to you from the side of a service provider?
I think like sir rightly mentioned about incident recording has been not a new phenomenon for at least people who have been in the IT industry. Recording cyber specific incident additionally has also started happening. However, we have tied back to the same ITIL framework that has been there historically followed. So enhanced AI is yet another tool which is landing up creating possibly an outcome. The outcome could be erroneous. It could be an event, incident, bias could be one of the situation that are arising. So as TSPs, individually while we have started doing this internally as we have adopted the journey of AI for us, these are recorded events. But one manner that it helps and supports in the framework as TEC has put across is, yes, it can be streamlined in a manner that the rest of the populace, if they have to refer by, can also be referred.
Because today there are no standalone companies, right? Every company is in the area. They are in the area of digital and IT. They are only doing. work in their own function. If you ask a bank, bank has to tell that I am an IT company in the service of doing banking. That is how it is changed. So IT plays a crucial role and AI will be a supporting arm in that. So this record keeping will make the ability for us to scale our AI and models as appropriately. With the advent that India wants to, and we have already announced three LLMs coming our way already, homegrown, home developed here, a platform like this will possibly help us manage and then refine our models well.
So you mentioned how enterprises are becoming digital first. And you also spoke in your initial remarks about AI for enterprises. So how do you look at this controlling costs of, you know, costs the infra part you did deal with, but how about the costs? Of AI for enterprises? Any thoughts on that?
Currently, it is still a significant amount that is being incurred upon. So the cost optimization, a larger chunk of cost optimization comes from the infrastructure as a whole. So about 80 -90 % of the cost to AI goes primarily on the infra in itself, both in store and as well in compute. The rest obviously comes in the skills. So today, while we definitely showcase the world that we have humongous talents that are getting built in the AI area, for an enterprise still to have these skilled engineers to build upon AI is still an adaptable work -in -progress area. So I think in the journey, we are now looking at AI to come in the aid of AI.
So we were in conversation with one of the AI -driven companies yesterday, and the way he highlighted back to us, telling that earlier… the total employee base was 10 ,000. Now there is a refinement and optimization by incorporating AI and thereby there is reduction in employee base. But then if we look at the people who are operating in AI, which was 30 now has gone to 3 ,000. So you cut down here and increase over there. So we were trying to tell them that the true power of AI is actually in making sure that AI is not touched with people, human. So reduction in human by upskilling that as appropriately is an important element for us to do. Thank you.
I’ll come to you, Dr. Upadhyay. You did, you know, I know I cut you off or sort of gave you a time pause there. Could you tell us a little bit more about what you were doing, what you’re doing with respect to disaster management, the application that you spoke about?
Disaster, yeah. Yeah. So disaster management, as you know, earlier, how did you? It used to happen. Suppose there is a cyclone in Odisha. A mail will go from IMD to chief secretary. Chief secretary will write to district collector. District collector would, in his best way, try to send the cyclone exactly to come. And we used to have thousands of lives lost, property lost. Today, using AI and the sensors, the system what we have done, this is one unified platform where all the alert generating agencies, IMD, CWC, FRI, DGSE, so all alert generating agencies are connected through APIs, auto. All the telecom operators are connected. All the alert dissemination agencies like SDMAs in the states are connected.
So it is all powerful one system. Now there is an alarm, a sensor alarm comes that there is a cyclone likely to, or rain likely. This is automatically read by the system. It prepares the message and finds out what is the geo -targeted area. Because earlier the problem was, they will put these kind of threats but nothing will happen. So people will take next time very casually. But today it is a geo -targeted system. It will alert only to the people who are in that belt. Suppose there is a cyclone hitting Gopalpur in Odisha it will only alert the people who are likely to be affected much before. And it will tell you also you need to evacuate you need to evacuate.
If you need to evacuate what is the arrangement by the government or you need to stay indoors. So all that happens and it was actually presented in parliament. The death in for example I am taking the case of Odisha where thousands of people died in 99. The death is zero. So what happened that after that we implemented because India is a large country and sometime a large population is to be alerted in some other cases. That time SMS gets delayed. You know SMS is a sequential process. SMS is sent by SMSCs. There was a new technology called cell broadcast where you don’t see the messages common. You don’t send through SMS you just broadcast it. So we developed a technology called Cell Broadcast And it was recently used in Cyclone Montha And how do you use AI?
Because now I am getting Inputs from various agencies I federate it My system federates all these information Using AI, builds one Particular message, finds out what is The right area where it is likely to hit And sends only to those people And the beauty of this system is, earlier there was a system Of group SMS, they will find people who are staying there Now even if you are a foreigner You are available at that particular time there It will pick your number and it will Give you the message. So tsunami Is coming, so We don’t know, people may be from here And there at the beach. So this has A very good, and in fact we have Published a paper in ITU, ITU has taken This as a report So this is going Forward, we feel that this particular System will meet The requirement of early warning for all By UN by 2027 And we are already talking to many countries And soon this solution will be Deployed in few countries which is Thank you.
Thank you.
In fact, in your presentation, you also spoke about fraud pro and so on. But I will, in the interest of time, I’ll move to Mr. Gorman about this fraud and scam. Now, you did in your opening remarks talk about the importance of collaboration across sectors. Also, the opportunities of, you know, of engendering innovation through collaboration for controlling or combating scams. Could you elaborate a bit upon that?
Sure. Thanks. Thanks for your question. I think this builds on actually the last couple of comments. I mean, what we’re talking about here is sharing data between multi -parties through standardized interfaces and then using AI or something or other to produce a good outcome. And all these things are innovations. They’re on the leading edge of something. If I start with the first thing, data sharing through standardized interfaces. standard interfaces, you know, GSMA has open gateway APIs program and that is contributing to providing data points which can be used in assessing risk for transactions. There’s other data that can be shared that could help address scams earlier in the cycle. Example, there’s lots of other data points to do and that’s the proof of concept that GSMA is working on in Southeast Asia, sharing data.
The challenge with doing that is you’re at the borders of regulatory compliance. You’re talking about private information or personal information or maybe not. There’s sometimes debate. But to be effective, you’re talking about being able to measure the risk on a particular individual user by sharing information across multiple parties. That requires some regulatory support, sandboxes or other activities to help find, to develop the innovation that finds the solutions that help combat scams. I think one of the things we need to focus on in industry is how do we create that nurturing environment that permits exploration of data sharing in a privacy -enhanced way. There’s lots of nice new technologies that have the impact while complying with the regulations and the privacy we want to maintain.
But ultimately, from a mobile operator point of view, I would say there’s four pillars in this combating scam thing. The one is the network, making sure the network cannot be manipulated in favor of the scammers. And that’s by CLI spoofing, all that sort of stuff. Let’s cut that out. If you introduce AI, there’s other things you can do on top. The second is what can mobile operators expose to the ecosystem so that the ecosystem can measure and respond to risk? Open gateway APIs is one thing. The POC I talked about before is another. And there may be other things. The third is what can mobile operators provide as services to their customers? in the same way the physical environment you can provide hard hats and things like that there’s things you can help customers and they can choose to acquire or choose to use them of their own choice to help protect them online and the fourth thing is digital skills digital skills historically we’ve considered is a destination in actual fact we now know we’re never going to hit that final point skills are going to continue to adopt and to adapt and it’s critical that we focus on all four pillars and that from a regulatory point of view and ecosystem point of view we’re collaborating so that the data can flow we can try and test things and we overcome the prejudice that may be stopping innovation because there’s an expectation that you can’t do these things so it requires policy makers regulatory to sponsor to nurture these things I mean I can guarantee I work into 90 % of mobile operators in Asia Pacific and I start a sentence with I want to suggest we use consumer data for I won’t get past halfway through the sentence they’ll say nope you can’t do that But in actual fact, if we want to be successful, no single entity, especially no mobile operator, has all the information.
I mean, if a mobile operator arbitrarily starts turning off SIM cards because they think maybe that traffic looks a bit dubious, I mean, you’ve only got to look at the Optus outage in Australia where three or four people died because they couldn’t call emergency services. You don’t want to be taking that action. It requires collaboration, regulatory support and policy support.
Yeah, thank you. Thank you. Network, ecosystem, hard hats and upskilling. I think that’s a good way to end the discussion here on the panel. But we have time for one question. Yes, Mr. Jha. We have less than two minutes.
Thank you. Very quick, very brief. The question from Mr. Julian Gorman. as we have said we are under attack may we attack anytime. We have also said that we should align with the global trends in order to combat these fraud and all those things. You have heard our panelists who are the icons in their field of manufacturing and standardization and PSPs. Could you suggest two steps that global leaders should take to align the world with themselves and two steps that India should take to align with the world. Thank you.
I mean two steps globally. So the proof of concept we are trying to do in Southeast Asia is actually prove that data can be shared domestically but across borders also in a safe secure way and has impact on controlling scams. One thing we need to remember with scams all we are doing by taking action against scams is increasing the cost of the business case and if we increase the cost of the business case here then another area becomes more favorable and that could be just different types of scams or it could be different locations and so that leads to the what do we need to do globally. We need to act across borders. We need to act as a collective global community.
GSMA has a program called United Against Scams there will be a lot of things about that in Barcelona but India is obviously taking great action domestically or taking steps domestically sharing that knowledge across borders and being able to share that data across borders is important and so I would leave it at those two points
Thank you, it also gives us pause for thought, maybe as regulators we also need to look at collaborating efforts across regulators because there are again sectoral issues that we need to do and so with that we are now at the end of the session, I would request the audience to give a big round of applause to my panelists who have given us very good insight into the topic at hand and thank you so much
Thank you moderator sir and all our distinguished panelists for such a vibrant discussion on usage of responsible AI the standards, the repository, various government app for enhancing consumer experience. Your insights will greatly benefit the overall digital ecosystem. Now I would request Dr. M .P. Tangirala to present mementos to our distinguished speakers as a token of appreciation. First to Mr. Julian Gorman. To Dr. Rajkumar Upadhaya. To Mr. Mathan Babu. To Mr. S .T. Abbas. Now I invite Sri A .K. Jha, Principal Advisor, TRAI to present a memento to the moderator of this session, Dr. M .P. Tangirala as a token of appreciation for moderating such a productive session Thank you so much, sir Now I take this opportunity to invite all the speakers for a group photograph I once again would request Chairman, sir, M .P.
Tangirala, Secretary, sir and all the Principal Advisors to please join the session speakers of this panel for a group photograph Thank you give a huge round of applause to all the panelists for joining us. Thank you. Thank you. Thank you. Thank you.
“The moderator introduced the Technology Security and Data Privacy Officer of Vodafone India and senior DDG S.T. Abbas as panelists.”
The knowledge base lists the Technology Security and Data Privacy Officer at Vodafone India and senior DDG S.T. Abbas as panel members, confirming their identification in the session [S100].
“AI must be deployed so as to minimise false‑positives and fully respect privacy and regulatory constraints.”
A related statement in the knowledge base stresses that false-positives should be kept very low when deploying AI for fraud detection, supporting the claim [S105].
“The session highlighted the need for enhanced collaboration among regulators across different sectors.”
The knowledge base notes that Dr Tangirala concluded the session by emphasizing the need for greater collaboration among regulators, providing additional context to the report’s emphasis on cross-sector cooperation [S1].
“The scam economy moves faster than regulation, prompting GSMA to create the Cross‑Sector Any‑Scam Task Force involving many organisations.”
The knowledge base reports large-scale scam activity (e.g., billions of spam instances and millions of scammers flagged), underscoring the magnitude of the problem that the task force aims to address [S54].
The panel displayed strong convergence on four main themes: collaborative ecosystems and shared standards; privacy‑by‑design and data protection; AI as a core tool for fraud/scam mitigation; and the need for standardized, cross‑border data sharing platforms. These points were repeatedly echoed across industry, standards and policy perspectives.
High consensus – the repeated alignment across diverse speakers indicates a unified direction for responsible AI deployment in telecom, suggesting that future policy and industry initiatives are likely to prioritize collaborative standards, privacy safeguards, and AI‑driven fraud prevention.
The panel broadly concurs that AI is vital for enhancing trust, fraud mitigation, and service quality in telecom. However, clear fault lines emerge around (i) the extent of human oversight versus full automation, and (ii) the preferred architecture for data governance—external, privacy‑preserving data sharing versus internal data lake consolidation. Additional nuanced tensions appear regarding voluntary incident reporting and the balance between privacy certification and innovation‑friendly regulation.
Moderate disagreement: while all participants share the overarching goal of trustworthy AI‑enabled telecom services, the divergent views on human control, data sharing models, and regulatory balance could lead to fragmented implementation strategies, potentially slowing coordinated progress on industry‑wide standards and trust‑building measures.
The discussion evolved from an introductory concern about AI accountability to a multi‑layered exploration of practical AI applications, infrastructural challenges, and governance mechanisms. Key comments—particularly those introducing cross‑sector collaboration, concrete AI success stories, the pitfalls of data silos, and the proposal of a voluntary incident‑reporting standard—served as turning points that redirected the conversation toward systemic solutions and policy frameworks. Julian Gorman’s coalition narrative and his four‑pillar model acted as the central pivot, linking technical implementations with regulatory and global coordination needs. Collectively, these insights shaped a cohesive narrative that balanced innovation with trust, underscoring the necessity of collaborative standards, centralized AI infrastructure, and continuous skill development to sustain responsible AI in telecom.
Disclaimer: This is not an official session record. DiploAI generates these resources from audiovisual recordings, and they are presented as-is, including potential errors. Due to logistical challenges, such as discrepancies in audio/video or transcripts, names may be misspelled. We strive for accuracy to the best of our ability.
Related event
