India’s telecom regulator has put artificial intelligence firmly on the rulebook. On October 30, the Telecom Regulatory Authority of India (TRAI) released draft guidelines on the responsible use of AI in telecom networks and services, and opened them for public comments until November 27, 2026. The message running through the draft is simple: if a telecom company uses AI, it answers for what that AI does, whether the system was built in-house, bought from a vendor, or put together from both.
The draft lands at a moment when AI is already woven deep into telecom operations. Operators use it to manage networks, predict faults, handle customer queries, detect fraud and plan capacity. TRAI’s argument is not that this should stop. It is that the risks that come with it need clear ownership and a paper trail.
What the draft asks operators to do
At the heart of the proposal is a board-approved internal AI policy. Every covered operator would have to adopt one, get it cleared by its board of directors or equivalent governing body, and file it with TRAI within seven working days of it taking effect, along with any later amendments. The policy would have to be reviewed at least once every financial year.
That policy must classify every AI system the operator uses as high risk, medium risk or low risk. The classification would rest on how significant the system is to network operations, what financial and operational outcomes it touches, and what it could mean for consumers. Nothing gets a free pass: the draft says no AI system may be used unless it appears in the operator’s inventory.
Operators would also have to keep that inventory current, covering active, inactive and decommissioned systems, with each entry recording the system’s source, purpose and risk class. Before any system is deployed, or changed, it would need validation and an assessment of accuracy, explainability, fairness, resilience to drift and human oversight. A designated officer would have to approve each system before it goes live.
Accountability stays with the operator
One of the draft’s sharpest provisions deals with outsourcing. Buying an AI system from a supplier does not move accountability along with the purchase. The operator remains answerable for the outcome, and the draft expects supplier agreements to carry enforceable provisions on compliance, audit and inspection, incident management and the allocation of responsibility. Suppliers would also have to disclose details of AI embedded in telecom equipment or systems.
Oversight would sit with a board-approved AI Governance and Safety Committee led by a senior executive. The committee would review performance, incident records, vulnerabilities and gaps in oversight at least once a year. The draft also calls for a central record of AI incidents and a separate central record of AI-related harm reported by telecom users, along with how each case was resolved.
How incidents would be reported
AI incidents would fall into four severity categories: critical, high, moderate and low. Critical and high-severity incidents would have to be reported to TRAI in a prescribed format, manner and timeframe. All incidents, whatever their severity, would have to be recorded and investigated.
There is one notable gap the draft itself leaves open. The exact format, manner and timeframe for reporting critical and high-severity incidents is left to a later prescription, so operators will be watching for that follow-up document. The draft also does not set out a separate penalty schedule for non-compliance, relying instead on TRAI’s existing regulatory mandate.
Who is covered, and what it builds on
The guidelines would apply to entities holding a licence or authorisation to provide telecom services or to establish, operate, maintain or expand networks, if they use or intend to use AI in their networks, services or both. That scope covers network and service operations, rather than every AI company that merely sells a model to the sector.
The draft builds on the government’s India AI Governance Guidelines, released in February 2026, and rests on seven guiding ideas, including trust, accountability and safety. TRAI has cited its authority under Section 11(1)(b)(v) of the TRAI Act, 1997, which covers quality of service standards and consumer protection.
What happens next
Stakeholders have until November 27, 2026 to send written comments to TRAI, addressed to Akhilesh Kumar Trivedi, Advisor (Networks, Spectrum and Licensing). Once comments come in, TRAI will weigh them before finalising the guidelines. For now, the draft is a consultation document, not a finished rulebook, and its final shape will depend on what stakeholders say before the deadline.
FAQs
What has TRAI proposed?
Draft guidelines for the responsible use of AI in telecom networks and services, released on October 30, 2026, requiring board-approved AI policies, risk classification, governance committees and incident reporting.
Who has to follow the proposed rules?
Licensed telecom operators and authorised network entities that use, or plan to use, AI in their networks or services.
How are AI systems classified under the draft?
Each system would be rated high, medium or low risk, based on its significance to network operations, financial and operational impact, and implications for consumers.
What happens if an AI incident occurs?
All incidents must be recorded and investigated. Critical and high-severity incidents must be reported to TRAI in a prescribed format and timeframe.
When can stakeholders comment?
Written comments are invited until November 27, 2026, addressed to TRAI Advisor (Networks, Spectrum and Licensing) Akhilesh Kumar Trivedi.
