TRAI Recommendations on Leveraging Artificial Intelligence and Big Data in the Telecommunication Sector (20 July 2023)

The National Digital Communications Policy 2018 (NDCP-2018) committed the Government to "leveraging Artificial Intelligence and Big Data in a synchronised and effective manner" to improve quality of service, spectrum management and network security. To build a roadmap, the Department of Telecommunications (DoT) wrote to TRAI on 6 June 2019 seeking recommendations on that provision (clause 2.2(g)).¹

TRAI's work took four years. It held a virtual conference on AI, machine learning and big data in the telecom sector on 5 and 6 August 2020, and workshops on the metaverse in September and December 2022. It issued a consultation paper titled "Leveraging Artificial Intelligence and Big Data in telecommunication sector" on 5 August 2022. Comments were originally due on 16 September 2022 and counter-comments on 30 September 2022, and both deadlines were extended to 14 and 28 October 2022. TRAI received comments from 22 stakeholders and one counter-comment, held an online Open House Discussion on 17 February 2023 and finalised the recommendations on 20 July 2023.¹²

TRAI noted that AI's impact is not limited to telecom, so it took "a holistic approach" covering all sectors. For that reason the framework it proposed "cannot be treated in isolation" and is a common one for every sector.¹ The recommendations also built on TRAI's July 2018 recommendations on privacy, security and ownership of data in the telecom sector, which it reiterated, and on its 18 November 2022 recommendations on the data economy, which had proposed a statutory Data Digitization and Monetization Council (DDMC).¹³

The document has four chapters: background; adoption of AI (definitions, risks, international experience, the case for regulation and access to data); enablers for adoption (telecom use cases, AI infrastructure, skills, accreditation, challenge programmes, model democratisation, MLOps and privacy-enhancing techniques); and a list of 21 recommendations.¹

Definition. TRAI declined to adopt a single definition, noting that there is no globally accepted one and that stakeholders cited NITI Aayog, ISO/IEC 22989, the OECD and the draft EU AI Act. It instead treated AI as a set of technologies "aimed at approximating some aspect of human cognition and learning" through machine learning and modern data techniques.¹

Risks. The analysis lists low-quality and biased data, inaccurate or biased algorithms, privacy and security failures, AI-enabled surveillance, deepfakes and disinformation, and unethical use.¹

Self-regulation. TRAI considered industry initiatives such as the IEEE ethics initiative and the Partnership on AI, and concluded that self-regulation "may lack accountability and enforcement mechanisms" and that such measures are not directly enforceable by government.¹

Regulating technology or use cases. Applying one set of mandatory requirements to all AI would restrict innovation, because not every startup could comply. TRAI therefore favoured classifying use cases by risk. Systems used for law enforcement, education, employment, credit and healthcare would be high risk and would have to comply with mandatory requirements before deployment. Limited-risk applications such as chatbots would face transparency requirements, and minimal-risk applications would face few or no rules.¹

Principles. It reproduced the five OECD and G20 AI principles (inclusive growth and well-being; human-centred values and fairness; transparency and explainability; robustness, security and safety; accountability) and proposed a suggestive matrix showing how they would apply differently to high-risk and low-risk use cases. AIDAI would finalise the matrix after public consultation, and the principles would apply at every stage of the lifecycle: design, development, validation, deployment, monitoring and refinement.¹

Recommendations 4.1 to 4.3 set out the institutional core. The framework should apply across sectors and comprise (a) an independent statutory authority, (b) a Multi Stakeholder Body acting as its advisory body, and (c) risk-based categorisation of use cases governed by broad principles of responsible AI. The authority should be established "immediately" and called the Artificial Intelligence and Data Authority of India (AIDAI).¹

AIDAI was effectively a merger of two ideas. In November 2022 TRAI had proposed a Data Digitization and Monetization Council to oversee data digitisation, sharing and monetisation and an overarching framework for ethical data use. Because that body was also expected to handle AI-related questions, TRAI now proposed to rename it AIDAI. TRAI warned that creating too many statutory authorities "creates confusion for the sector" and suggested that AIDAI's work could be entrusted to TRAI itself, with suitable amendments to the TRAI Act.¹³

The proposed functions fall into three groups:

The Multi Stakeholder Body would include representatives of DoT, the Ministry of Information and Broadcasting, MeitY, DPIIT, the Department of Science and Technology and the Ministry of Home Affairs, two academics, four industry members, one legal expert in AI, one cybersecurity expert and up to two others. TRAI also asked that all AIDAI regulatory decisions be preceded by open consultation, and that MeitY be designated the administrative ministry for AI.¹

TRAI additionally recommended that India collaborate with other governments to form a global agency as the primary international body for the development, standardisation and responsible use of AI, and that India play a leading role.¹

Beyond the regulatory architecture, the recommendations are practical and directed mainly at DoT:¹

• Connection limits and security. Use AI/ML to analyse data in real time and verify the number of mobile connections held against DoT's limits, and deploy AI/ML-based tools in DoT's Telecom Security Operation Centre to generate network threat alerts.
• Grievances. Flow data from operators' redressal systems into DoT's Public Grievance Unit dashboards for AI/ML-based analysis.
• Unsolicited commercial communication. Amend the Telegraph Act, Telegraph Rules and Unified Licence conditions so that TRAI can direct licensees to run continuous machine-based identification and monitoring of bulk peer-to-peer messages, using pattern recognition, anomaly detection, traffic analysis and reputation analysis to counter phishing, spam and scams.
• Indigenous AI. Encourage AI for areas such as radio-frequency optimisation, beamforming, network slicing and edge computing, and, in broadcasting, speech recognition, automated video processing and content analysis and moderation, possibly funded through the Telecom Technology Development Fund. DoT should work with IISc Bangalore, IIT Madras, IIT Kanpur and others on indigenous use cases.
• Infrastructure and skills. Set up at least one AI Centre of Excellence in every State and Union Territory with access to bandwidth, compute and datasets and linked to 5G/6G labs; form a committee with MeitY, the Ministry of Education and AICTE on AI skills; and ask AICTE to mandate a course on the ethical use of AI. DoT's training institute NTIPRIT should offer courses for government employees on the iGOT platform.
• Challenge and bounty programmes under the Digital Communication Innovation Square scheme.

On accreditation, stakeholders were divided. Some thought accreditation of AI products was premature and could create entry barriers, while others wanted an optional mark comparable to the BIS ISI mark. TRAI concluded that AIDAI should coordinate with the Telecom Engineering Centre on accrediting laboratories that test AI products, and that sector regulators should be involved in sector-specific accreditation.¹

Stakeholder comments in the document show a divided response. Several respondents argued that a comprehensive data protection law should be passed before any data-sharing framework for AI was created, and that the absence of guidelines was a blind spot for industry. Others considered that no regulatory intervention or sector-specific compliance was needed at that stage and that, at most, a cross-sectoral multi-stakeholder body should later refine voluntary principles.¹

TRAI's own analysis acknowledged the alternative of leaving AI to sectoral regulators. It cited NITI Aayog's 2021 approach paper, which said that India's existing sectoral regulators could continue to oversee AI-led innovation in their domains "for the time being". TRAI nevertheless concluded that a horizontal framework was preferable, because many AI risks, including biased algorithms and human-machine relationships, fall outside any existing regulator's jurisdiction.¹⁴

The document is a set of recommendations to the Government and does not itself bind anyone. It was for the Union Government to accept, modify or ignore them.

As of 15 October 2026 the sources reviewed for this entry did not identify any statute or notification establishing an AI and Data Authority of India or amending the TRAI Act to create an AI regulator.

The Government has instead chosen a different route. The India AI Governance Guidelines released in November 2025 rely on a review of existing laws rather than a standalone AI statute and propose an AI Governance Group, a Technology and Policy Expert Committee and an AI Safety Institute.⁵ On 13 April 2026 MeitY constituted the AI Governance and Economic Group (AIGEG), chaired by the Union Minister for Electronics and IT, as a central coordinating body. MediaNama noted that its membership of ministers and senior officials omits the independent regulators, including TRAI, that the guidelines had recommended, and tied this to the Government's repeated stance against AI-specific legislation.⁶