Quick Summary
- India's telecom regulator TRAI released draft guidelines on the responsible use of artificial intelligence in telecom networks and services on October 30.
- The proposal covers accountability, risk classification, human oversight and safety assessments, with stakeholder comments invited until November 27, 2026.
- The draft is a consultation proposal, not a final regulation, and its eventual impact will depend on the final wording and implementation requirements.
A sector-specific framework for AI deployment
The Telecom Regulatory Authority of India has proposed a framework for how artificial intelligence should be used in telecommunications networks and services. The draft guidelines, released on October 30, aim to apply the broader principles of India’s AI governance framework to a sector that supports communications, digital services, commerce and critical infrastructure. TRAI has opened the document for stakeholder feedback, with comments due by November 27, 2026.
The proposal reflects a shift in AI governance from general principles toward rules for particular industries. Telecom operators increasingly use automated systems in network management, customer support, fraud detection and service providey. These systems can improve efficiency, but errors may affect connectivity, billing, customer access or the resilience of infrastructure. The regulator’s central concern is that organisations remain accountable for the outcomes of AI systems they deploy, including systems developed by external vendors.
According to reporting on the draft, telecom entities would need to classify AI systems according to risk, establish governance arrangements and maintain human oversight for important decisions. A risk-based approach aims to distinguish a low-impact automated function from a system that could affect network operations or customers at scale. The final framework will need to define how those categories are assigned and what controls apply to each one.
Accountability becomes an operational requirement
The proposal is significant because telecom operators cannot avoid responsibility simply by purchasing an AI product from a third party. A vendor may build the model or software, but the operator controls how it is integrated into its network and customer processes. That means the operator needs to understand what the system does, what data it uses, how it behaves when conditions change and how a human can intervene when the output is unsafe or incorrect.
Risk assessments and inventories can help organisations answer those questions. An inventory should identify where AI is used, the business or network function it supports, the data it processes and the teams responsible for it. A risk assessment can then consider potential effects on service availability, privacy, security, fairness and consumer outcomes. Such documentation is useful only if it remains current as systems are updated or connected to new data sources.
Incident reporting is another important part of the proposed approach. If an AI system contributes to a serious failure, the operator needs a process for identifying the event, preserving evidence, limiting harm and informing the regulator. This is particularly relevant when automated systems are involved in network configuration or security operations, where an error can propagate quickly across connected services.
TRAI has not yet finalised the guidelines. Telecom companies, technology vendors, consumer groups and other stakeholders have an opportunity to examine the draft and raise concerns about feasibility, definitions and compliance costs. The consultation process matters because broad requirements can be difficult to implement unless they are translated into measurable controls and clear timelines.
Operators will be watching how the regulator defines high-risk AI, what constitutes a reportable incident and whether requirements differ by network function or company size. Vendors may seek clarity about the information operators must obtain from suppliers. Consumers, meanwhile, have an interest in ensuring that automated systems do not make consequential decisions without an effective route for review.
The proposal signals that AI governance is becoming part of telecom operational compliance rather than a separate innovation exercise. The final rules could influence procurement, model testing, internal audit and incident response across the sector. Until the consultation closes and a final version is issued, organisations should treat the document as a draft framework and avoid presenting its proposals as obligations already in force.
The draft still has to become an enforceable framework
The consultation stage is important because telecom operators, equipment providers and consumer groups may have different views about what constitutes an acceptable risk. A rule that is too broad could make routine network operations harder, while a rule that is too narrow may leave gaps around automated decision systems, customer information and service reliability. Operators will need to understand how the proposed controls would fit with existing security, privacy and incident-reporting obligations.
The practical test will be whether the final guidelines define responsibilities clearly enough for a network operator to act during an incident. That includes identifying who must be notified, what information has to be preserved and how quickly a report must be submitted. Until TRAI completes the consultation and publishes a final version, the document should be treated as a draft proposal rather than a rule that is already in force.