The Telecom Regulatory Authority of India, known as TRAI, released draft guidelines on 30 October 2026 for the responsible use of artificial intelligence in telecom networks and services. Stakeholders can send written comments by 27 November 2026. The draft is a consultation document, so its proposals can change before any final version is notified. The regulator has not presented the draft as a set of final rules already in force.
The full title of the draft is Guidelines on Responsible Use of Artificial Intelligence in Telecommunication Networks and Provision of Telecommunication Services. It applies to entities that hold a telecom licence or authorisation and use, or plan to use, artificial intelligence in their networks or services. TRAI said the work builds on the India AI Governance Guidelines released by the central government in February 2026. Those guidelines set out principles for the use of artificial intelligence across sectors, and the telecom draft applies them to a sector that carries voice calls, data traffic, emergency services and business systems.
Telecom operators already use artificial intelligence for network management, traffic optimisation, predictive maintenance and customer service. The draft asks how those systems should be governed, checked before use and watched after deployment. It also asks who remains responsible when a system is bought from a supplier rather than built inside the company.
What TRAI proposes for telecom AI policies
Under the draft, each covered entity would adopt an internal artificial intelligence policy approved by its board of directors or an equivalent governing body. The policy would sort every artificial intelligence system into high risk, medium risk or low risk. Classification would rest on the importance of the system for network operations, its financial and operational effects, and its possible effect on consumers. No system could be deployed without approval from a designated officer, according to the draft.
Entities would file their policies, and any later amendments, with TRAI within seven working days of the policy coming into effect. They would review the policy at least once in each fiscal year. The draft also asks for an inventory of active, inactive and decommissioned artificial intelligence systems. Suppliers would face due diligence checks before procurement, and systems would need validation, conformity and impact assessments before deployment or modification.
The draft also proposes a governance and safety committee led by a senior officer for artificial intelligence governance and safety.
How TRAI would handle AI incidents and accountability
The draft proposes four incident severity levels: critical, high, moderate and low. Entities would keep central records of incidents, investigate failures and report critical and high severity artificial intelligence incidents to TRAI. The form and time frame for that reporting would be prescribed under the final guidelines. Entities would also record and resolve harm reported by users that is linked to an artificial intelligence system.
Accountability is the sharpest part of the draft. The draft proposes that the authorised entity remains accountable for the outcome of all artificial intelligence systems it uses, whether the systems are developed internally, procured or outsourced from a supplier, or built through a mix of those routes. That point matters for telecom networks because many systems arrive as part of equipment, software licences or managed services, and the operator still runs the network that customers rely on.
The draft is built around seven guiding principles: trust, people-centricity, innovation, fairness and equity, accountability, safety, and resilience and sustainability. Those principles come from the February 2026 governance guidelines. Contracts would also need to cover matters such as audit rights, incidents and accountability.
What TRAI consultation means for operators and users
For consumers, the draft does not announce any change to call rates, data tariffs, SIM rules or recharge plans. Its effect, if adopted in its current form, would sit behind the service, in how operators approve systems, test them and respond when they fail.
Not every reader of the draft treats it as a binding rule. Satya N. Gupta, a former principal adviser at TRAI, told Mint that the guidelines are more in the nature of guidance than mandates or directives, and fall under light touch regulation. That reading fits one feature of the draft noted in coverage: the document uses firm language in places, but it does not set out a separate penalty schedule for non-compliance. TRAI has issued the proposals under its existing regulatory mandate rather than as a separate artificial intelligence law for telecom.
Comments can be sent electronically to Akhilesh Kumar Trivedi, Adviser for Networks, Spectrum and Licensing at TRAI, at advmn@trai.gov.in. Telecom operators, technology vendors, civil society organisations and academic institutions are among those invited to respond. The regulator will consider the responses before it decides the final form of the guidelines, including the details of serious incident reporting and the duties that apply to each covered entity.