The Indian government is examining whether social media platforms should continue to enjoy blanket legal protection for user-generated content when their algorithms determine what users see, recommend or suppress. The discussions could lead to changes in Section 79 of the Information Technology (IT) Act, 2000, potentially exposing platforms to greater liability for content distributed through their recommendation systems.
The issue is whether a platform that actively influences the reach of a post can be treated in the same way as an intermediary that only transmits information between users. Social media companies do not create most of the content published on their services, but their algorithms decide which posts appear in feeds, which videos are recommended and which accounts gain visibility.
“Any social media intermediary that uses algorithms to decide the ranking, placement, promotion or suppression of user-generated content is no longer just a pipeline to ferry such content but has an active role in deciding its fate. A blanket safe-harbour protection against liability for such content, therefore, needs a relook,” the official said in a statement to Business Standard as reported here.
What Section 79 protects, and what could change
Section 79 provides legal protection to internet intermediaries against liability for third-party information, data or communication hosted or transmitted through their platforms. This protection is subject to conditions, including that the intermediary does not initiate the transmission, select its recipient or modify the information being transmitted.
Intermediaries must also exercise due diligence and comply with government-prescribed rules. The protection can be withdrawn if a platform abets, aids or induces an unlawful act involving the content. It can also lose the exemption if it fails to remove unlawful material after receiving a government notice or court order.
The proposed rethink centres on how these conditions apply to social media platforms that use automated systems to determine content visibility. Under the existing framework, platforms can argue that the content originates with users and that their role is to provide the infrastructure through which it is shared.
However, recommendation systems do more than carry posts from one user to another. They assess user activity, content characteristics and engagement patterns to decide what appears in individual feeds. This can affect the reach of a post well beyond the audience that follows the account publishing it.
The government is examining whether such intervention should affect the legal protection available to platforms. According to an official quoted in Business Standard report, only platforms that enable peer-to-peer messaging and do not interfere with the content delivered from one user to another could remain exempt under the proposed approach.
When recommendation becomes a form of editorial control
The debate over safe harbour raises a question about the distinction between hosting content and distributing it. A social media platform may not write a post, produce a video or create an image, but its recommendation system can determine how many people encounter it.
Most large social media platforms use algorithms to personalise feeds and recommend content. These systems take into account factors such as user interactions, viewing patterns and the type of content a person engages with. The objective is to show users material that they are more likely to view or interact with.
This gives platforms control over content distribution. A post published by an account with a limited following can reach a much larger audience if the recommendation system picks it up. The same post may remain largely unseen if the algorithm does not recommend it.
The distinction becomes relevant when the content is unlawful or harmful. If a platform’s algorithm recommends a misleading video, promotes an abusive post or distributes a deepfake to a large audience, questions arise over whether the company should be held responsible for the role its systems played in that distribution.
However, treating every recommendation as an act of publication could create a difficult legal standard. Algorithms rank and recommend large volumes of content through automated processes. A platform’s decision to recommend a post does not necessarily mean that it has reviewed or endorsed every claim made in it.
The distinction between automated distribution and editorial approval will therefore be important in determining the scope of any proposed liability. The government will also have to establish how much control over content distribution should be enough to affect safe-harbour protection.
What it could mean for content moderation
A change in the safe-harbour framework could affect how platforms approach content moderation. Companies may have to reassess the way their recommendation systems identify, rank and distribute content that could violate Indian laws.
This could lead to greater scrutiny of content that is promoted through recommendation feeds, rather than only material reported by users or flagged by authorities. Platforms could also face pressure to explain how their systems handle certain categories of content and what steps they take to prevent unlawful material from gaining wider reach.
There is a risk, however, that greater liability could encourage platforms to take a more cautious approach to content distribution. Companies may choose to restrict the reach of material that could attract legal scrutiny, even when its legality is not clear.
It could also affect how platforms handle disputes over content. If companies face greater legal exposure for recommending material, they may be more likely to remove or restrict content when its legality is contested, rather than wait for a formal determination.
For publishers, this raises questions about the control social media companies exercise over the distribution of news. Publishers produce and verify their content, but platforms decide how it is presented to users and how much visibility it receives through recommendation systems.
The same issue applies to creators who depend on platform algorithms for audience growth and advertising income. Changes to recommendation policies could affect the reach of their content and, in turn, their ability to generate revenue.
How other countries have approached platform responsibility
The question of whether algorithmic recommendations should affect platform liability has also come up in other jurisdictions, particularly the US and the European Union.
In the US, the question came before the Supreme Court in a case from 2023 involving YouTube’s recommendation algorithms. The family of a woman killed in the 2015 Paris terror attacks alleged that YouTube’s recommendation systems helped Islamic State (ISIS) spread propaganda and recruit supporters. They argued that Google should not receive protection under Section 230 of the Communications Decency Act, 1996, for content promoted through its recommendation systems.
The Supreme Court did not rule on whether algorithmic recommendations fall outside Section 230 protection. It sent the case back for reconsideration after ruling in a related matter that the allegations did not establish that the platforms had knowingly provided substantial assistance to ISIS. The question of liability for recommended content therefore remained unresolved.
The European Union has taken a different approach through its Digital Services Act (DSA). The law requires online platforms to explain the main parameters used by their recommendation systems and allow users to modify or influence them. Very large online platforms must also offer at least one recommendation option that does not rely on user profiling. They are also required to assess and address risks arising from their services, including the spread of illegal content and threats to minors.
An earlier ruling by the Court of Justice of the European Union in 2021 also examined the distinction between hosting content and actively participating in its distribution. The court held that automatically indexing uploaded material, providing search functions and recommending videos were not, by themselves, enough to establish that platforms had an active role giving them knowledge of or control over illegal content.
These approaches show that algorithmic recommendations can attract regulatory scrutiny without automatically making platforms liable for every piece of content they distribute. The legal treatment depends on the nature of the platform’s involvement and the obligations imposed by the relevant framework.
What India needs to clarify
For India, the central question will be how to distinguish between platforms that host user-generated content and those that actively influence its distribution. If the use of recommendation algorithms itself becomes grounds for withdrawing safe-harbour protection, the scope of the proposed changes could extend to most major social media services.
The government will need to clarify whether liability would apply to all forms of algorithmic ranking and recommendations or only where platforms promote specific categories of unlawful content. It will also have to establish how automated recommendations, knowledge of illegal content and deliberate promotion would be treated under the amended law.
There are also implications for freedom of expression and the availability of content online. If platforms face legal action over material recommended by their systems, they may restrict content more aggressively to avoid liability. This could affect legitimate news, political commentary and other forms of user-generated content.
