India's telecom regulator drafts AI rules for carriers, making operators answerable for vendor-built systems

The Telecom Regulatory Authority of India (TRAI) published draft guidelines on October 30 on the responsible use of artificial intelligence in telecommunication networks and services, opening a consultation that runs until November 27. Titled the Guidelines on Responsible Use of Artificial Intelligence in Telecommunication Networks and Provision of Telecommunication Services, the draft is the telecom sector's first dedicated AI framework since India released its national AI governance guidelines in February 2026, and it applies to licensed entities that use or plan to use AI in networks or services.

Board-approved policy and three-tier risk classes

Each covered entity would have to adopt an internal AI policy approved by its board of directors or equivalent governing body, setting out how AI systems are classified, approved, monitored and governed, file the policy with TRAI within seven working days of it taking effect, and review it at least once a year. A board-approved AI Governance and Safety Committee, headed by a senior executive designated as AI Governance and Safety Officer and drawing members from network operations, service delivery, IT, data management, regulatory and consumer affairs, would oversee the framework. AI systems would be classified as high, medium or low risk according to their materiality to network operations, service delivery, financial and operational outcomes and consumers, and could be used only after approval by a designated officer, with validation, conformity and impact assessments before deployment or modification and an inventory of active, inactive and decommissioned systems.

Accountability for vendor systems; four-tier incident reporting

The sharpest provision is accountability: the applicable entity would be accountable for the outcome of every AI system it uses, whether developed internally, procured from a third party or a combination of both, with contracts dividing costs or indemnities but not the accountability itself. All AI incidents would have to be recorded, investigated and analysed, classified into four severity bands of critical, high, moderate and low, with the critical and high bands reported to TRAI in a prescribed format and timeframe; human oversight of all AI systems would be mandatory. The draft is still a consultation document, not a final rule: former TRAI principal adviser Satya N. Gupta told Mint the guidelines are closer to a code of conduct under light-touch regulation than to enforceable mandates, and the final text will follow TRAI's review of submissions.