The Meta episode: Why India’s IT platform rules need a right to restoration

The Parliamentary Standing Committee on Communications and Information Technology gave Mark Zuckerberg 3 days to tender an unqualified personal apology for the restriction of the Prime Minister's Facebook video between roughly 12:30 AM and 5 AM on July 28, 2026. Failing that, it recommended that the Union government consider withdrawing Meta's protection under Section 79 of the Information Technology Act, 2000.

Most of the commentary since has asked whether the threat is executable. It is not; that can be disposed of briefly. The more useful question has gone largely unexamined. What does an intermediary owe a user whose lawful content it restricts and is that obligation enforceable by anyone at present? On the current text and practice, close to nothing. That is the defect this episode has exposed and no apology addresses it.

Section 79(1) exempts an intermediary from liability for third party information, subject to sub-sections (2) and (3). Sub-section (2) conditions the exemption on the intermediary not initiating the transmission, not selecting the receiver, not selecting or modifying the information and observing due diligence. Sub-section (3) disapplies it where the intermediary conspired, abetted, aided or induced the unlawful act, or failed to remove material expeditiously upon actual knowledge, which Shreya Singhal v. Union of India confined to a court order or a government notification.

Nothing in that structure permits a parliamentary committee to extinguish the immunity. A standing committee may summon persons, take evidence and report to the House with recommendations. It exercises scrutiny, not adjudication. Whether a particular intermediary forfeited protection in relation to particular content arises in litigation, is answered instance by instance, and is answered by a court. A platform-wide withdrawal would require amendment of the Act and the government's reported decision to seek a legal opinion suggests the difficulty has registered.

There is a constitutional overlay as well. In Kunal Kamra v. Union of India, the Bombay High Court held the fact check unit amendment to the 2021 IT Rules unconstitutional and beyond the rule-making power conferred by the parent Act. That amendment had made safe harbour contingent on an executive body's view of what was false or misleading about government business. The appeal is pending. The proposition that immunity may be conditioned on satisfying the executive has been litigated once already and has not fared well.

Turn now to the part of the framework that ought to have operated here and did not.

Rule 4(8) of the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 applies where a significant social media intermediary disables access to information on its own initiative. It must first notify the user with a notice explaining the action and the grounds for it. It must provide an adequate and reasonable opportunity to dispute the action and request reinstatement. And the resident grievance officer must maintain oversight over the resolution of such disputes. Three obligations, then. Reasons, an opportunity to contest and supervised resolution.

Consider what a user actually receives. The Kerala journalist KA Shaji reported that a post carrying an image of Rahul Gandhi and Shashi Tharoor with student protesters was restricted in India. The notice said that access had been restricted through an automated system implemented pursuant to legal requirements under the IT Rules. Shaji removed the students and the placards, leaving only the two politicians, but the restriction recurred. In May 2026, the Congress made a comparable allegation about photographs of Rahul Gandhi with the actor Vijay, again attributing the block to IT Rules compliance.

That notice does not identify which sub-clause of Rule 3(1)(b) was engaged. It does not say whether the restriction followed a direction under Section 69A, a court order, a private complaint, or the platform's own classifier. It does not indicate whether any human considered the matter. It recites the existence of Indian law as though the recitation were itself a ground.

That is not a reason. A reason is a statement which permits the person affected to know what has been alleged and to answer it. A notice asserting only that some legal requirement somewhere applies renders the second limb of Rule 4(8) formally available and practically void. One cannot dispute an unstated ground.

The point sharpens against Section 69A practice. Directions under that provision issue through a designated officer, on specified grounds, following a committee process. Rule 16 of the 2009 Rules imposes confidentiality on them. Whatever the merits of that, the direction exists as an identifiable instrument, capable in principle of being pleaded and tested. The generic notice collapses the distinction. A restriction originating in a statutory order and one originating in a probabilistic classifier appear identically to the user and the platform need not say which it was.

Two consequences follow.

First, it determines the remedy. If a Section 69A direction exists, the aggrieved party challenges the direction and the intermediary is a compliant executor. If none exists, the restriction is private governance by an entity claiming intermediary status and the route is Rule 4(8) read with the grievance mechanism under Rule 3(2) and escalation to a grievance appellate committee under Rule 3A. A notice concealing which situation arises defeats the choice of remedy at the threshold. That gap is already in litigation, X having told the courts that Section 79(3)(b) is being used to construct a parallel blocking process bypassing the safeguards attached to Section 69A.

Second, it bears on safe harbour itself. Section 79(2)(b) requires that the intermediary not select the receiver of the transmission. An intermediary which decides on its own classification - without a legal instrument - that an item shall be unavailable to Indian users performs an act of selection the sub-section contemplates as inconsistent with the exemption. The government has advanced a version of this argument about algorithmic curation and paid amplification. The over-blocking cases illustrate it more cleanly, because the act is discrete, attributable and admitted.

The irony is worth stating. The strongest available argument that Meta behaves as something other than a passive conduit comes not from the Prime Minister's video, but from the routine restriction of ordinary users' political posts, which has drawn no regulatory attention at all.

The BBC Eye investigation published on July 3, 2026 found Instagram carrying paid advertisements in India which used explicit search terms and routed users to Telegram channels selling child sexual abuse material. MeitY required the ads disabled and an explanation of how they had been approved. In a controlled test during the 2024 general election, researchers from Ekō and India Civil Watch International submitted 22 AI manipulated political advertisements. 14 were approved within 24 hours.

Safe harbour analysis has limited work to do here. Section 79 protects an intermediary in respect of third party information which it hosts or makes available. An advertisement is a placement the platform reviews under its own policies, prices, accepts payment for and targets to an audience it selects. The selection of the receiver is the service being sold. The statutory language sits awkwardly with the proposition that such a placement is third party information for which the platform bears no responsibility.

Rule 4(4) applies with obvious force. It requires significant social media intermediaries to endeavour to deploy technology-based measures to proactively identify child sexual abuse material, with human oversight and periodic review of those tools for accuracy and fairness. A finding that advertisements bearing explicit search terms cleared review is prima facie evidence that the review obligation was not discharged and that is determinable on documents the platform must maintain.

The committee asked for an apology and obtained one. 3 demands would have produced more.

1. Give Rule 4(8) content, by amendment or advisory. Any geo-restriction applied to Indian users should state on its face whether it rests on a legal instrument, identified by number and issuing authority, or on the intermediary's own assessment. In the latter case, the sub-clause of Rule 3(1)(b) relied upon. A platform invoking Indian law should be able to name the law.

2. Establish a restoration obligation with a clock. The present scheme obliges acknowledgement of grievances within 24 hours and disposal within 15 days, with shorter timelines prescribed for certain removal requests. 15 days is an eternity for political speech during an agitation. Wrongful restriction of lawful speech should carry a defined restoration timeline and a record.

3. Require aggregate disclosure. False positive rates for political content by Indian language, the proportion of automated actions reversed on appeal and median time to human review, published periodically and independently audited. Rule 4(4) already requires periodic evaluation of automated tools for accuracy and fairness. Nothing presently requires the results to be produced to anyone, which renders the obligation unfalsifiable.

Meta's answer to the episode has been to promise additional layers of senior review for the Prime Minister and other prominent verified accounts. A bespoke escalation lane for the eminent is an admission of the defect rather than a cure. The law's concern is not with whose speech was restricted. It is with whether anyone was told why.

The Meta episode exposes a structural defect in India's platform-governance architecture: the law specifies duties of moderation, but gives the affected speaker no equally concrete mechanism for obtaining reasons, measuring error, or restoring lawful speech before its value has disappeared.

Prasanth Raju is an Advocate and Counsel in Chambers of Prasanth Raju, Bombay High Court.