On 30 September 2026, the Telecom Regulatory Authority of India (TRAI) held an open house on its consultation on quality of service (QoS) in 5G networks, including network slicing. The Broadband India Forum (BIF), the Consumer Care Society and MediaNama's Nikhil Pahwa argued that slicing must not dilute net neutrality. Airtel and Reliance Jio objected to two of TRAI's proposals: a 21-day advance notice for new slices and an 80% Physical Resource Block (PRB) utilisation threshold (MediaNama). Both camps are partly right. The regulation that results should protect general internet quality and leave operators room to sell genuinely different network products.
What TRAI has proposed
The consultation paper is dated 5 August 2026. As reported, it would require operators to inform TRAI at least 21 days before launching a new 5G slice. Each slice would be treated as a separate tariff offering with its own QoS monitoring. Radio resource utilisation during peak hours should not exceed 80%, and no more than 1% of cells in a licensed service area should cross that level (Indian Infrastructure; Telecom Lead). Operators would also have to publish measured speeds for slices that offer enhanced mobile broadband.
The strongest case for the sceptics
The worry is not irrational. Slicing carves a virtual network out of shared radio capacity and guarantees it certain performance. If a slice can be sold as "better quality for one application", it starts to look like the paid prioritisation India has already banned. BIF's Debashish Bhattacharya said slicing can coexist with net neutrality only if it is "transparent and non-discriminatory" and does not harm general internet services for ordinary users. BIF's position is that application-specific quality on demand would breach the rules. Pahwa warned that content-specific slices would turn operators into "gatekeepers instead of gateways" (MediaNama). India's framework gives these concerns real force. TRAI's Prohibition of Discriminatory Tariffs for Data Services Regulations, 2016 bar tariffs that discriminate on the basis of content, source, destination or application (TRAI). In 2018 the Department of Telecommunications (DoT) wrote the principle of non-discriminatory treatment of content into telecom licences (Business Standard).
Why a blanket ban on slicing would also be wrong
The same framework anticipated this debate. The licence amendments exempt "specialised services", defined as services other than internet access that are optimised for specific content, protocols or user equipment where optimisation is necessary to meet specific QoS requirements. Those services must not be detrimental to the availability and overall quality of internet access service, and must not be offered as a replacement for it (DoT; MediaNama, 2018). A slice for remote surgery, industrial automation or emergency communications is a specialised service in the plainest sense. Even the Rajkot consumer group at the open house backed emergency-service prioritisation, provided rural and ordinary consumers are not adversely affected (MediaNama).
A pro-innovation reading is therefore that enterprise and mission-critical slices are lawful and valuable. 5G's economics depend partly on such products, and a rule that treated every differentiated service as a neutrality violation would push that investment elsewhere. The legal test should be whether a slice is defined by objective technical parameters, such as latency, throughput and reliability, or by the identity of a particular app or content provider. The Consumer Care Society made this point when it asked that differentiation rest on "objectively defined parameters".
The operators' objections: one is stronger than the other
Airtel's Tarun Chitkara said the 80% threshold has "no consumer impact" and would force capacity upgrades unnecessarily. Jio's Udai Kumar Srivastava called PRB utilisation an internal radio resource management parameter and the rule "technically inappropriate" (MediaNama). The technical objection has substance. A utilisation ceiling is an input measure. Regulation that works better, and costs less, targets outcomes such as measured speeds, latency, call and session success, and complaint rates. TRAI already proposes publishing slice-level speed data, and that is the kind of evidence consumers and competitors can use. If TRAI keeps a capacity trigger, it should justify it with data linking high PRB use to degraded user experience.
The objection to the 21-day notice is weaker. Notification is not approval. It costs operators a short delay and gives the regulator a chance to spot a slice that would harm general internet access before it launches, instead of after consumers are affected. A slice that is truly specialised can be described in a notice in a few pages. If three weeks is too long for competitive reasons, a shorter fixed window, or a deemed-clearance mechanism for slices that fit pre-published templates, would answer the concern. Scrapping notice would leave the net neutrality safeguards without any way of being enforced.
What a proportionate outcome looks like
- Keep disclosure, drop input mandates. Require filing of slice parameters and publication of measured performance. Replace or justify the 80% PRB rule with outcome-based benchmarks.
- Draw the line at applications. State plainly that slices may differ by technical performance class and customer segment, but may not be sold on the basis of favouring named content or apps over general internet access.
- Protect the baseline. Tie the rules to the specialised-services condition in the licence: no slice may degrade general internet access, and none may be marketed as a substitute for it.
- Settle the venue. MediaNama reports that a separate net neutrality review is under consideration at DoT. A QoS regulation should not quietly become the place where net neutrality is rewritten.
The open house showed that neither camp wants to give up either innovation or neutrality. The details of the draft decide whether slicing makes networks better for everyone or merely a premium for some. Transparent, outcome-based and application-neutral rules deliver the first without risking the second.