Sectoral regulators are now being asked to apply national AI governance principles. Using TRAI's draft guidelines on AI in telecom, discuss the opportunities and challenges.

The India AI Governance Guidelines (Feb 2026) set out seven sutras for every sector. On 30 October 2026, TRAI released draft guidelines that apply these sutras to AI used in telecom networks and in the provision of telecom services [1]. This is one of the first cases of a sectoral regulator turning national principles into sector rules. It is a promising model, but it shows the gap between principles and rules that can be enforced.

Opportunities

• Rules that fit the sector: National sutras are general. A sector regulator can turn "Safety, Resilience & Sustainability" into norms for live networks, and Part 4 of the framework gives practical guidance [1].
• Builds on existing practice: TRAI had already directed operators to deploy an AI/ML-based UCC_Detect system to catch spam [2]. That scale demands automation: 31.09 lakh spam complaints were filed in 2025 [3].
• Legitimacy through consultation: The draft is open for public comments until 27.11.2026, which makes the final rules more predictable and more widely accepted [1].
• Supports innovation: "Innovation over Restraint" matches TRAI's stated risk-based approach and its preference for sandbox testing [4].
• A model for other regulators: It could serve as a template for other sectoral regulators applying the national framework.

Challenges

• Soft law in a high-stakes sector: "Guidelines" are not binding, yet the draft itself says telecom carries critical information infrastructure and emergency services, where an AI failure could hurt national security [1].
• No safe place to test: TRAI recommended a regulatory sandbox on 12.04.2024, but DoT sent it back for reconsideration through a back-reference in March 2025 [5]. Without it, AI goes straight onto the live network.
• Explainability and redress: If an algorithm wrongly flags a genuine caller as spam, that caller can be cut off. "Understandable by Design" needs clear reasons for such decisions and a way to appeal them.
• Technology alone has limits: Despite AI-led enforcement, 89,936 repeat offenders needed six-month restrictions in 2025 [3].

Sectoral application makes national AI principles practical, but only if it comes with enforcement. TRAI should set binding regulations for high-risk uses such as network fault management and fraud detection, keep guidelines for low-risk uses, and DoT should settle the sandbox back-reference quickly. That would give real meaning to the first sutra, "Trust is the Foundation", and would advance SDG 9 on resilient infrastructure.