India is building a more stringent legal and technological defence against deepfakes as artificial intelligence makes fabricated voices, images and videos cheaper to produce, harder to detect and capable of spreading across digital platforms within minutes.

The emerging framework combines faster content removal, mandatory labelling of synthetic media, traceable metadata, platform accountability, indigenous detection tools and public awareness. In a written response submitted to the Lok Sabha, Union Minister of State for Electronics and Information Technology Jitin Prasada said the objective was to maintain an “open, safe, trusted and accountable” cyberspace while protecting citizens from impersonation, fraud, privacy violations and manipulated content.

The government has approved 13 Responsible AI projects at educational institutions under the Safe and Trusted AI pillar of the IndiaAI Mission. These include projects devoted to detecting manipulated videos and cloned voices—two forms of synthetic content increasingly associated with financial fraud, reputational attacks and disinformation.

Among the initiatives is Saakshya, a multi-agent deepfake detection framework being developed by IIT Jodhpur and IIT Madras. Other projects include AI Vishleshak, which is designed to strengthen audio-visual forgery detection, and a real-time voice deepfake detection system being developed by IIT Kharagpur. The programmes reflect an important shift from relying primarily on foreign detection products towards building tools suited to India’s linguistic and technological environment.

The urgency is evident. Generative AI systems can now recreate a person’s appearance and voice from limited samples. Such material can be used to circulate false political statements, imitate senior executives, manipulate public opinion, perpetrate investment scams or create non-consensual intimate imagery.

CERT-In has classified the threat as serious, warning that deepfakes can cause misinformation, financial loss, fraud and lasting reputational damage. Its November 2024 advisory carried a “high” severity rating. CERT-In’s advisory also highlighted their use in social-engineering attacks.

India’s response rests on several overlapping statutes. The Information Technology Act, 2000 covers offences such as identity theft, cheating through impersonation, privacy violations and the electronic transmission of obscene or sexually explicit material. Section 69A provides for blocking access to specified information, while intermediaries that fail to observe their legal obligations risk losing the conditional immunity available under Section 79.

Deepfake-related conduct can also attract provisions of the Bharatiya Nyaya Sanhita, 2023. Section 353 addresses false information, rumours or reports capable of causing fear or public mischief, while organised cybercrime involving synthetic material may be prosecuted under Section 111. Depending on the content and intent, other laws concerning child protection, defamation, sexual offences or national security may also become applicable.

The most consequential intervention came through the amended Information Technology Rules notified on February 10, 2026 and brought into force on February 20. The rules define synthetically generated information as artificially created or altered audio, visual or audio-visual material that appears authentic and may be indistinguishable from a real person or event. Pure text is not included within this specific definition, although text-based content remains subject to the broader laws governing unlawful information.

MeitY’s official clarification also excludes routine, good-faith editing such as colour correction, compression, transcription or noise reduction when it does not distort the meaning of the original material.

Platforms providing tools for creating or modifying synthetic media must deploy reasonable technical safeguards to prevent the generation or circulation of unlawful content. Prohibited categories expressly include child sexual exploitation material, non-consensual intimate imagery, false electronic records and deceptive impersonations.

Synthetic content that is lawful must be prominently labelled. Audio material must carry an audible disclosure, while images and videos must display a visible marker. Platforms must, to the extent technically feasible, embed permanent metadata or another provenance mechanism, including a unique identifier connecting the content to the system used to create it. They cannot enable users to suppress or remove these identifiers. These requirements are intended to give citizens a basic signal that apparently realistic content has been artificially produced.

The amendments also recognise the speed at which online harm can escalate. Once an intermediary receives a court order or a valid, reasoned government direction concerning unlawful information, the removal deadline has been reduced from 36 hours to three hours.

The period for addressing specified removal complaints has fallen from 72 hours to 36 hours. In sensitive cases involving nudity, sexual content, impersonation or artificially morphed images, the response time has been cut from 24 hours to two hours. The general grievance-disposal period has also been shortened from 15 days to seven days.

Users dissatisfied with a platform’s grievance officer can approach a Grievance Appellate Committee through the government’s online portal. The mechanism is significant because it gives individuals a route of appeal against content-moderation decisions without immediately requiring court proceedings.

Large social-media intermediaries—those with at least five million registered users in India—face additional duties. These include appointing India-based compliance officers, publishing regular transparency reports and deploying automated tools against specified unlawful content. Messaging services may also be required, under legally prescribed conditions, to help identify the first originator of content connected with serious offences.

The technological response is being accompanied by public education. Under the Information Security Education and Awareness programme, 6,650 workshops have reached more than 11.37 lakh participants, including students, teachers, police personnel, government employees and members of the public. Multilingual handbooks, videos, posters and children’s stories are also being distributed to improve cyber hygiene.

The government’s strategy signals that deepfake regulation cannot depend on a single law or detection tool. Effective protection will require coordination among platforms, investigators, courts, researchers and citizens. Detection systems must continually evolve because generative models are improving rapidly, while labelling and metadata standards will work only if platforms implement them consistently.

India has now created a framework that places responsibility across the AI-content chain—from the system generating synthetic media and the user publishing it to the platform distributing it. Its success, however, will ultimately be judged by enforcement: how quickly harmful deepfakes are identified, whether victims receive meaningful relief and whether safeguards can curb deception without restricting legitimate creativity, satire, research and innovation.

Join the community of 2M+ industry professionals.

Subscribe to Newsletter to get latest insights & analysis in your inbox.

All about ETCISO industry right on your smartphone!