By Shivanand Pandit
In an effort to curb online misinformation, the Ministry of Electronics and Information Technology (MeitY) has proposed new rules to regulate fake or misleading content. The draft rules, released on October 22, 2025, say that Artificial Intelligence (AI) and social media platforms must clearly label content created by AI. The government is seeking feedback from the public and industry on these rules until November 6, 2025.
The draft framework brings legal clarity by defining “synthetically generated information” for the first time under Indian law. It places AI-generated content within the ambit of existing takedown obligations, ensuring that misleading or harmful material can be addressed promptly. In terms of labelling, the rules specify that images and videos must carry visible labels covering at least 10% of the display area, while audio content must be identified as synthetic during the first 10% of playback. In other words, the label should cover 10% of the content’s area and apply to all types of AI-generated content, including text, video, and audio—not just realistic-looking images. These measures are designed to make it immediately apparent to users when content is artificially generated.
Enforcement responsibility is divided between two groups: the companies that create AI tools or synthetic content and the platforms that host user-generated material. Platforms will be required to ask users to declare whether their uploads are synthetically generated, employ automated detection technologies to verify authenticity, and act on content flagged through grievance redressal mechanisms. This dual approach is intended to balance innovation in AI with protection against potential harms, fostering accountability across India’s rapidly expanding digital ecosystem.
Timing meets wisdom
For too long, the proliferation of synthetically generated content has gone unchecked, eroding trust in digital spaces and exploiting the openness of social media. The recent surge of manipulated videos targeting senior ministers underscores the urgency of the problem. These are not harmless pranks but deliberate digital forgeries aimed at deceiving citizens and committing fraud. Finance Minister Nirmala Sitharaman has warned that we are entering an era of “hacking trust,” emphasizing the widespread misuse to which technology can be put. In this context, the government’s move to formally define and regulate synthetic content under the Information Technology Rules, 2021, is a critical step. By proposing mandatory labelling and metadata for AI-generated media, it has signalled a commitment to transparency while maintaining space for innovation. Users must be able to distinguish between real and artificial content at a glance, and platforms must be accountable for ensuring that clarity.
Certifying authentic content, rather than flagging fakes, could be a more efficient approach, reducing the constant monitoring burden. Yet in India, where the digital user base is enormous and the consequences of a viral deepfake could be severe, the government’s approach appears pragmatic. Whether the focus is on authenticating genuine content or marking artificial content, the objective remains the same: to prevent deception from being mistaken for truth. These nuances can be debated and refined during the consultation process.
Equally significant is the government’s simultaneous effort to strengthen the framework for online content takedowns. By restricting the power to senior officials—joint secretaries and above, or deputy inspector generals and above—the government has introduced a level of accountability that was previously absent. The provision for a review by secretary-rank officers further ensures that takedown orders are proportionate and not arbitrary. These reforms follow the Karnataka High Court’s affirmation of the government’s authority in such matters, making the timing particularly notable. Rather than using the court’s endorsement to push through draconian measures, the government has opted to formalize and refine the procedures.
Nevertheless, the challenge cannot be addressed by the government alone. Deepfakes represent a rapidly evolving technological frontier that outpaces regulatory responses. Tools for creating synthetic content are increasingly accessible, while the platforms that host them are global and often slow to act. In this dynamic, enforcement will always be reactive. This is why regulation must be paired with awareness and shared responsibility. Labelling is not a panacea but one of several tools in combating misinformation and manipulation. Platforms must enhance detection systems, and users must cultivate a habit of questioning the authenticity of what they see and share. Technology’s promise should not be hindered by fear, nor should its misuse be ignored in the name of progress. The government’s proposals aim to strike this delicate balance. Ultimately, success will depend not on the number of rules written but on the vigilance with which they are implemented and observed by all stakeholders, including everyday users, who remain the primary targets of such scams.
Key Red Flags
While the dangers posed by deepfakes—such as non-consensual intimate imagery and election manipulation—are undeniable, the current regulatory proposals risk overreach. They threaten to impose excessive censorship, compel speech, and introduce invasive monitoring that could suppress lawful online expression. The draft rules apply to any content “algorithmically created, generated, modified or altered… in a manner that… appears… authentic or true.” This broad definition could easily encompass satire, remixes, parodies, or harmless edits, making the regulation virtually universal in scope.
Under the draft, platforms and tools that enable creation or editing would be forced to embed permanent identifiers and display visible or audible labels covering “at least 10 percent” of any work, regardless of context, with no provision for removal. This constitutes compelled speech and mirrors intrusive censorship mechanisms reminiscent of cinema and OTT content regulation. Such measures carry a high risk of collateral censorship while doing little to deter bad actors, who can simply ignore the rules.
The draft also ties “synthetically generated information” into due diligence and traceability requirements, heightening privacy and encryption concerns. Merely labelling or embedding metadata in AI-generated content is unlikely to address the underlying harms. Technically, these obligations are straightforward to write into rules but extremely difficult to implement reliably—and trivial to bypass. A nontechnical user can generate AI content and remove watermarks, labels, or disclaimers within minutes. Moreover, the requirement that AI disclaimers or labels cover at least 10 percent of a work will disrupt user experience and slow page load times, inadvertently pushing users toward foreign AI platforms that do not comply with India’s IT rules.
While the compliance burden falls heaviest on significant social media intermediaries, smaller platforms—even those with fewer than five million users—may face disproportionate costs. Effectively labelling synthetic content is challenging given the sheer volume of digital material produced today, much of which is AI-assisted. For smaller intermediaries, the obligation applies only to content generated within their apps, leaving third-party AI-generated uploads unlabelled. Ambiguities in the definition of synthetic media further complicate compliance, while the prevalence of AI tools in everyday content creation may lead to routine, excessive labelling, diluting the effectiveness of notices over time. Automated detection systems remain unreliable: lightly edited or AI-assisted content may be misclassified, while sophisticated deepfakes may evade detection entirely.
Unlike mandatory smoking warnings in cinemas and on OTT platforms, which were imposed with limited industry enthusiasm and little evidence of impact, content labelling has been embraced by major social media and AI companies from the outset as synthetic content became more accessible. Meta, Inc., for instance, began labeling AI-generated content on Facebook last year, while the Coalition for Content Provenance and Authenticity (C2PA) has brought together industry players to address “digital provenance” issues, borrowing a concept from the art appraisal world. Using subordinate legislation may not have been the optimal way to implement this change — the IT Rules already govern streaming services, social media content removals, and the now-prohibited real-money gaming sector, all without prior parliamentary scrutiny. Eventually, these rules need to be deliberated by elected representatives. Policymakers have long noted that regulation often trails behind technological innovation. The government must therefore be prepared to follow up this proposal with agile measures—both by relaxing outdated rules and introducing new ones as circumstances demand.
The ongoing consultation period presents a critical opportunity to address these gaps. Policymakers should consider a broader range of interventions—including tackling text-based misinformation—without resorting to heavy-handed measures that stifle AI adoption and innovation. India’s proactive stance is commendable, but the complexity of AI demands precision, nuance, and technical feasibility. The real solution lies not only in regulation but in building public digital literacy and robust institutions of media trust—essential defenses against the erosion of factual consensus in an era of increasingly sophisticated synthetic media. The current draft is only a starting point; the challenge now is to develop a legislative framework that balances innovation, free expression, and public safety.
