Deepfake Laws: How India, EU, China, and the US Manage Synthetic Media
As of mid-2026, global governments have moved from experimental policy to strict mandatory frameworks for managing synthetically generated information.
The Shift Toward Mandatory Regulation
The regulatory landscape for deepfakes and synthetic media has shifted significantly by mid-2026. Governments across the globe have transitioned from non-binding guidelines to enforcement-heavy frameworks, aiming to curb malicious use while addressing the rapid proliferation of artificially generated content. While the approaches differ, the shared goal is to ensure transparency through mandatory disclosure and accountability for intermediaries.
India: IT Rules (Amendment) 2026
India formalized its approach to synthetic media with the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, which took effect on February 20, 2026. This framework introduces specific obligations for platforms hosting AI-generated content.
Under the new rules, the government has defined Synthetically Generated Information (SGI), which includes any audio, visual, or audio-visual content altered by computer resources to appear authentic. Platforms are now required to label SGI clearly and, where technically possible, embed provenance metadata. Perhaps most significantly, the rules mandate a 3-hour takedown window for content flagged by courts or authorities, with a stricter 2-hour deadline for non-consensual deepfake nudity.
The European Union: Transparency Under the AI Act
The European Union has prioritized transparency as the cornerstone of its AI regulation. The transparency obligations of the EU AI Act, detailed in Article 50, apply from August 2, 2026. Providers and deployers of AI systems must disclose when content is artificially generated or manipulated.
The EU framework requires content to be marked in a machine-readable format to ensure accountability at scale. While there is a grace period until December 2, 2026, for systems already on the market before the August deadline, the rules establish a rigorous standard for all synthetic media. Notably, these requirements apply broadly, with limited exceptions only for specific law enforcement activities or artistic works where transparency might disrupt the user experience.
China: Standardized Labelling Requirements
China adopted a highly structured approach to synthetic content through the mandatory national standard GB 45438-2025, which became effective on September 1, 2025. This standard, titled Cybersecurity Technology – Labeling Method for Content Generated by Artificial Intelligence, requires both visible and invisible markers.
Service providers must apply explicit labels, such as on-screen text or audio indicators, while also embedding implicit metadata within files. The responsibility extends to social media platforms, which are tasked with verifying that the providers they host have implemented these labelling mechanisms. The law also strictly prohibits the tampering, deletion, or forging of these identifiers.
United States: The Federal and State Hybrid Model
The United States employs a tiered regulatory strategy. At the federal level, the TAKE IT DOWN Act (Tools to Address Known Exploitation by Immobilizing Technological Deepfakes on Websites and Networks Act), signed on May 19, 2025, mandates that platforms follow specific notice-and-takedown procedures regarding non-consensual intimate imagery and creates new federal criminal prohibitions.
Complementing this, a state-level patchwork has emerged, with 46 states enacting laws by spring 2026 focused on election-related deception and synthetic intimate imagery. Furthermore, states including New York, California, Illinois, and Tennessee have expanded right-of-publicity statutes to protect individuals against the unauthorized use of their digital replicas, moving beyond traditional commercial usage requirements.
| Jurisdiction | Core Regulatory Instrument | Enforcement Focus |
|---|---|---|
| India | IT Rules Amendment 2026 | Takedown timelines & SGI labelling |
| EU | EU AI Act (Article 50) | Machine-readable transparency |
| China | GB 45438-2025 | Mandatory explicit/implicit labels |
| US | TAKE IT DOWN Act / State Laws | Criminal prohibitions & publicity rights |
Frequently Asked Questions
What constitutes Synthetically Generated Information in India?
Under Rule 2(1)(wa) of the 2026 Amendment, it covers any audio, visual, or audio-visual content created or altered by computer resources to appear indistinguishable from real-world events or people.
When do the EU transparency obligations take full effect?
While the AI Act applies from August 2, 2026, generative AI systems already on the market have until December 2, 2026, to implement required machine-readable markings.
Do US federal laws cover all types of deepfakes?
The current federal landscape, specifically the TAKE IT DOWN Act, focuses primarily on non-consensual intimate imagery, while other areas like election interference remain primarily within the jurisdiction of individual states.
Sources
- Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 [MeitY]
- The EU AI Act Transparency Obligations [European Union]
- GB 45438-2025: Cybersecurity Technology – Labeling Method for Content Generated by Artificial Intelligence [China Standards]
- TAKE IT DOWN Act Summary [Federal Law]