TRAI drafts AI accountability rules for India's telecom operators

India's telecom regulator has given operators and other licensed network providers until 27 November to respond to draft guidelines that would make them fully accountable for every AI system they use, whether built in-house, bought from a third party, or a mix of the two.

The Telecom Regulatory Authority of India (TRAI) opened the consultation this week on a proposed framework for the responsible use of artificial intelligence in the provision of telecoms services, a scope that explicitly includes AI deployed inside network infrastructure. The regulator framed the move as a step toward what it called the "consistent and responsible implementation" of national AI governance guidelines issued by the government in February.

The proposal places concrete obligations on licensees. Companies would have to conduct risk assessments of any AI technology they use, define and register AI-related incidents, and report incidents deemed of critical severity to TRAI in a prescribed form and within a set timeframe.

What does the framework require in practice?

TRAI proposes that operators and registered network providers define internal policies on their use of AI and appoint a committee, led by a senior executive, to verify those policies are actually being implemented. Each company would have to produce an internal document covering several defined areas:

• risk assessments of AI systems;
• defining and registering incidents;
• ensuring human oversight of AI operations;
• data governance; and
• a consumer protection and grievance redressal mechanism.

The framework's legal reach is broad. TRAI states the guidelines are intended for any company holding a telecoms licence or any entity authorised to operate and maintain networks under the Telecommunications Act 2023, the legislation that replaced India's century-old Indian Telegraph Act as the sector's governing statute.

The accountability clause is the sharpest edge of the draft. Licensed companies "will be deemed accountable for the outcome of all AI systems used by it, irrespective of whether the AI systems are developed internally, procured/outsourced from third party or a combination thereof". In plain terms, an operator cannot shift liability onto an AI vendor.

Why is TRAI acting now?

The regulator set out its reasoning in the consultation document. "While the integration of AI into telecommunications is likely to significantly enhance efficiency, it may also introduce new risks if due diligence is not exercised before deployment of AI systems or if due supervision is not exercised over autonomous learning and systemic interactions," TRAI explained.

The risk catalogue it flagged is specific. TRAI pointed to inadvertent biases, unintended model behaviour, and what it termed "adversarial attacks", which it asserted could affect the resilience of telecoms infrastructure.

What does this change for Indian operators?

For now, nothing is binding. The document is a draft, and the consultation window runs to 27 November, during which interested parties can submit feedback. If adopted in its current shape, the framework would translate the February national AI governance guidelines into enforceable sector-specific practice: executive-led oversight committees, documented risk processes, human supervision requirements, and a duty to report critical AI incidents to the regulator on a defined clock.

The burden would fall hardest on operators leaning heavily on third-party AI tools, since outsourcing development would not outsource accountability. Stakeholder responses filed before the 27 November deadline will determine how far the final rules soften or sharpen those obligations.