India’s Telecom Regulator Proposes Responsible AI Rules for Operators

Indian telecom operators would need a board-approved AI policy, a register of every AI system they run and a duty to report serious AI failures to the regulator, under draft TRAI AI guidelines published on 30 October. The Telecom Regulatory Authority of India (TRAI) has put the draft out for consultation, with written comments due by 27 November 2026. Nothing in it applies yet.

The proposal tries to turn India’s broad AI principles into duties for one critical sector. In a press release, TRAI said it “has decided to frame guidelines on responsible use of AI in telecommunication networks and provision of telecommunication services after following stakeholders’ consultation”. The draft builds on the India AI Governance Guidelines of February 2026, and says AI in telecoms brings risks such as inadvertent bias, model drift and adversarial attacks that could harm the resilience of telecom infrastructure.

Who the TRAI AI guidelines would cover

The draft would apply to any company holding a licence under the Telegraph Act, 1885, or an authorisation under the Telecommunications Act, 2023, to provide telecom services or run telecom networks, if it uses or plans to use AI in those networks or services. That reaches mobile and broadband operators and the companies building network infrastructure, a sector that is spending heavily on new capacity, as Tata Communications’ 10,000km fibre plan shows.

Each operator would write an internal AI policy, approved by its board or an equivalent body, and file it with TRAI within seven working days of it taking effect. The policy would sort every AI system into high, medium or low risk, based on how much it matters to the network and services, including its effect on financial and operational outcomes and on consumers. Using any AI system would need sign-off from a designated officer.

The policy would also have to cover human control. In TRAI’s words: “All AI systems shall be subject to an appropriate human oversight for risk mitigation as per a predefined mechanism.” The draft adds that the level of oversight should follow each system’s risk rating.

How operators would track AI incidents

Under the draft, every AI incident would be classed as critical, high, moderate or low severity and recorded in a central database, using a Telecommunication Engineering Centre standard, TEC 57090:2025, or an equivalent taxonomy. Incidents would be handled under a set procedure and investigated, and critical and high-severity incidents, including material malfunctions and recurring failures, would be reported to TRAI “in a prescribed form, manner, and timeframe”. The draft leaves those details for later.

Operators would keep an inventory of all active, inactive and retired AI systems. “No AI system shall be used unless it is a part of the inventory,” the draft says. New or changed systems would face a conformity assessment covering accuracy, robustness, explainability, fairness, drift resilience and human oversight before deployment. Bought-in AI would need supplier due diligence and contracts that give the operator audit and inspection rights. A senior executive, the AI Governance and Safety Officer, would head a committee reviewing how deployed systems perform at least once a financial year, and would have to ensure fixes, including modifying or decommissioning systems where necessary.

Board sign-off puts telecom AI on the compliance map

For anyone who has worked in financial compliance, the shape is familiar. A board-approved policy, a risk-rated inventory, an incident register, supplier due diligence and a named senior officer are the building blocks of model risk management in banking. TRAI is proposing to carry that discipline into networks, where an AI fault could disrupt critical infrastructure rather than misprice a loan.

The most consequential clause may be the simplest. Operators “shall be 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”. That closes the obvious escape route of blaming a vendor, and it gives operators a reason to demand the disclosures and audit rights the draft describes. Equipment makers selling AI-enabled kit into India could face tougher contracts.

What is missing matters as much. The draft sets no penalties and no reporting deadlines, and leaves the “reasonable timeframe” for resolving AI harms reported by users undefined. Its human oversight test, “appropriate” oversight “as per a predefined mechanism”, lets each operator define what counts. Those gaps are likely targets for industry comments.

Some operators are already moving. SK Telecom gave its AI agents employee IDs and human supervisors, and vendors such as Calix are shipping AI models built for telecom networks. India’s approach contrasts with the UK, where regulators said existing law already applies to AI agents and set out no new sector rules.

What comes next for the draft

Comments go to TRAI’s Advisor for Networks, Spectrum and Licensing at [email protected] by 27 November 2026, and TRAI says submissions will be posted on its website. Only after that will the regulator settle the final TRAI AI guidelines, including the reporting format and timeframes that the draft leaves blank.

Sources

Primary:

• TRAI, “Draft Guidelines on Responsible Use of Artificial Intelligence (AI) in Telecommunication Networks and Provision of Telecommunication Services”, consultation draft, 30 October 2026 (PDF)
• TRAI, Press Release No. 145/2026, “TRAI releases draft Guidelines on Responsible Use of Artificial Intelligence (AI) in Telecommunication Networks and Provision of Telecommunication Services for stakeholders comments”, 30 October 2026

Press:

• None.

Related stories

• SK Telecom Gives Its AI Agents Employee IDs and Human Supervisors
• Calix Releases Open-Weight Telecom AI Model and Plans Gems for 2027
• UK Regulators Say Existing Law Already Applies to AI Agents