This note is the second segment of a two-part post on the effect of India’s (i) data protection regime – comprising the Digital Personal Data Protection Act, 2023 and the Digital Personal Data Protection Rules, 2025 (together, the “DPDP Framework”), and (ii) intermediary framework, respectively, on artificial intelligence (“AI”). The Indian legal framework governing intermediaries include the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 (“Intermediary Amendment Rules”), which amended the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 (“Intermediary Guidelines”) under the Information Technology Act, 2000 (“IT Act”).

Part I, published earlier as a separate note, examined the implications of the DPDP Framework for AI developers and deployers, including with respect to (i) consent and purpose limitation, (ii) data minimization, (iii) lawful purpose, (iv) rights of access, correction and erasure, (v) obligations of significant data fiduciaries, and (vi) children’s data. This Part II examines the impact of the Intermediary Amendment Rules on AI and its implications for investors, along with emerging business opportunities, and concludes the two-part post.

Impact of the Intermediary Amendment Rules on AI

Among other things, the Intermediary Amendment Rules define and regulate synthetically generated information (“SGI”). Intermediaries must use technical measures and automated tools to prevent the creation and dissemination of prohibited SGI and ensure labeling and metadata for permissible SGI.^[1] Compliance and takedown timelines under the Intermediary Guidelines have also been significantly shortened. In general, the Intermediary Amendment Rules are poised to have a major impact on AI-enabled platforms, including those that allow users to create and post content.

Previously, in October 2025, the Ministry of Electronics and Information Technology (MeitY) had issued proposed amendments to the Intermediary Guidelines (“Proposed Amendment”), including with respect to mandatory labeling and metadata obligations for AI-generated content. Compared to the Proposed Amendment (which primarily focused on AI labeling requirements), the Intermediary Amendment Rules focus on preventing deepfakes/ misinformation and a rapid response regime. In that regard, the Intermediary Amendment Rules introduce critical definitions to the Intermediary Guidelines for the purpose of addressing nuances in ‘synthetic’ (i.e., artificially generated) media.^[2]

Scope and Definitions

Audio, visual, or audio-visual information (“AV Content”) is broadly defined to include any image, photograph, graphic, video, sound recording, or moving visual recording, regardless of whether it was created or altered via a computer resource. While the Proposed Amendment had defined SGI broadly (thereby raising significant stakeholder concern),^[3] the Intermediary Amendment Rules have narrowed such definitional scope and set a higher threshold. SGI is now defined as artificially or algorithmically created AV Content (e.g., deepfakes) that appears real, authentic, or true, such that it portrays individuals or events in a manner ‘indistinguishable from a natural person or real-world event’. Further, it has been clarified that any reference to ‘information’ in the context of unlawful acts will include SGI, ensuring that synthetic content is subject to the same – and in some cases stricter – legal scrutiny as ‘traditional’ content.

To prevent the overregulation of benign AI applications, the Intermediary Amendment Rules provide specific safe harbor protections under Section 79 of the IT Act. Unlike the Proposed Amendment, the new rules include standard SGI exclusions for: (i) routine/ good-faith editing (e.g., enhancements such as color adjustment, noise reduction, or transcription that do not materially alter the substance or meaning); (ii) accessibility tools (e.g., when AI is used solely for translation, subtitles, or improving discoverability without manipulating material parts of the content); and (iii) professional/ educational material, such as the creation of templates, PDF files, or research outputs – provided they do not result in false documents or electronic records.

Enhanced Due Diligence Obligations

The Intermediary Amendment Rules significantly tighten the due diligence requirements for all intermediaries to ensure higher user awareness and swifter enforcement. Intermediaries must now inform users of their obligations, the platform’s right to terminate access for non-compliance, and the legal consequences of violations at least once every three months (as opposed to the previous annual requirement). Such user notification must be made available in simple language in English or any other language specified in the Eighth Schedule to the Constitution of India.

Specialized SGI Warnings

Intermediaries that offer tools enabling the creation or modification of SGI (such as AI image generators or voice cloners) must prominently provide additional warnings. Users must be cautioned that directing a computer resource to create unlawful SGI may attract penalties under the IT Act, the Bharatiya Nyaya Sanhita, 2023 (“BNS”), and other specialized laws like the Protection of Children from Sexual Offences Act, 2012 (POCSO).

Action Against Violating Users

Upon becoming aware of SGI-related violations, intermediaries are required to take ‘expeditious and appropriate action’, which includes: (i) immediate removal or disabling of access; (ii) suspension or termination of user accounts; (iii) evidence preservation (ensuring that account actions do not ‘vitiate the evidence’ required for legal proceedings); and (iv) identification and disclosure of the violating user’s identity to the victim or authorities in accordance with the law.

Strict Takedown and Redressal Timelines

The Intermediary Amendment Rules introduce stringent compliance timelines to mitigate spreading of harmful synthetic content as follows:

• Actual Knowledge: When an intermediary receives a court order or a reasoned intimation from a Government-authorized officer regarding unlawful content, it must remove or disable access within three hours (previously, 36 hours).
• Sensitive Content: Grievances related to non-consensual nudity, sexual acts, or impersonation (including artificially morphed images) must be addressed within two hours (as opposed to 24 hours).
• Standard Grievances: General user complaints must be resolved within seven days (previously, 15 days).
• Access Removal Grievances: Specific complaints regarding the removal of information under Rule 3(1)(b) of the Intermediary Guidelines must be resolved within 36 hours.

Operational Safeguards for Rapid Takedowns

To facilitate the 3-hour Government takedown window, the Intermediary Amendment Rules allow for one or more authorized officers (not below the rank of Deputy Inspector General of Police) to issue intimations. This change ensures 24/7 operational readiness and prevents bottlenecks in responding to viral misinformation or public order threats.

Dedicated SGI Due Diligence Framework

The Intermediary Amendment Rules create a dedicated regulatory framework for intermediaries which facilitate SGI, as follows:

• Proactive Prevention: Intermediaries must deploy reasonable and appropriate technical measures, including automated tools, to prevent users from creating or disseminating prohibited SGI. The prohibited categories are (i) sexually explicit content (including child sexual abuse material), non-consensual intimate imagery, and content invasive of bodily privacy); (ii) false records (e., SGI that results in the creation of false documents or electronic records, e.g., forged IDs or government letters); (iii) hazardous information (i.e., content relating to the development or procurement of explosives, arms, or ammunition); and (iv) deceptive impersonation (i.e., SGI that falsely portrays a natural person or real-world event via misrepresentation of voice, identity, or conduct in a manner likely to deceive).
• Labeling and Provenance: For permissible SGI, intermediaries must ensure the following: (i) prominent labeling (visual content must have a clearly noticeable label; audio content must have a prominently prefixed audio disclosure); (ii) metadata embedding (SGI must be embedded with permanent metadata and a unique identifier to enable the identification of the computer resource used for its creation; (iii) anti-tampering (intermediaries are prohibited from enabling modification, suppression, or removal of such labels or metadata.

Additional Obligations for Significant Social Media Intermediaries

Significant Social Media Intermediaries (“SSMIs”) are subject to enhanced ex-ante duties, as follows:

• Pre-publication Verification: Before allowing the publication of any information, an SSMI must require a user declaration regarding whether the content is SGI.
• Technical Verification: SSMIs must deploy technical measures to verify the accuracy of such declarations (rather than relying solely on user input).
• Mandatory Labeling: If confirmed as SGI via declaration or verification, such content must be clearly and prominently labeled prior to display.
• Shift from Endeavor to Mandatory: Rule 4(4) of the Intermediary Guidelines has been amended to remove language related to the endeavor to deploy – now explicitly requiring SSMIs to deploy appropriate technical measures to proactively identify rape, child sexual abuse, or identical previously removed content.

Legal Protections and Consequences

• Safe Harbor Protection: The Intermediary Amendment Rules clarify that an intermediary’s safe harbor protection under the IT Act remains intact when it removes content (including SGI) in compliance with such rules or through proactive technical measures. This clarification provides legal certainty to intermediaries performing good-faith removals of harmful synthetic content.
• Non-Observance: Failure to observe changes introduced by the Intermediary Amendment Rules to the Intermediary Guidelines will result in the loss of safe harbor. In such cases, the intermediary becomes liable for punishment under any law in force, including the IT Act and the BNS.

Key Takeaways

AI and AI-enabled platforms may be significantly impacted by the Intermediary Amendment Rules, and intermediary compliance is likely to: (i) require substantial technical modifications, (ii) affect user experiences, and (iii) consume significant resources.

AI developers that generate or modify content may now have to incorporate tamper-proof watermarking and labeling at the point of creation. When AI models are accessed by API providers, assistive tools, and third-party integration environments, such labeling must persist, requiring advanced engineering modifications in software. Developers may also need to build or license verification mechanisms to support intermediary obligations, accelerating the demand for auditability, provenance tracking, and forensic metadata technologies.

From a deployment perspective, labeling requirements may reshape risk assessments for products that embed generative AI features. Deployers and developers may need to assess whether and to what extent their AI-powered features fall within the scope of the Intermediary Amendment Rules.

SSMIs may need to implement/ adopt: (i) user interface elements that prompt SGI declarations at the point of upload, (ii) detection tools to cross-verify such declarations, and (iii) audit trails documenting such declarations and verifications. The platform architecture must be capable of preserving AV content labels.

DPDP Framework and Intermediary Amendment Rules: Implications for AI Investors

Strong data governance mechanisms, data provenance and SGI handling, and privacy-by-design capabilities are likely to become core indicators of scalability and defensibility, including for investors evaluating AI-focused targets. The DPDP Framework introduces compliance risks that must be assessed as part of the overall diligence process,^[4] given the significant penalties involved (up to INR 25 billion, depending on the nature of violation and other factors). Legal consequences arising from the Intermediary Amendment Rules may also prove material.

Rigorous Due Diligence

Investors must conduct comprehensive due diligence to ensure a target has the necessary rights to collect and use personal data for its AI models. This includes verifying that the target has obtained valid consents, complied with all data protection laws, and established robust governance and security measures to prevent data breaches. In addition, investors need to seek warranties on compliance with data protection regulations and indemnities in case of material breaches of the DPDP Framework.

Compliance with the dedicated SGI due diligence framework under the Intermediary Amendment Rules, as applicable, will also require careful attention, along with takedown, redressal, verification, and labeling capabilities.

Contractual and Insurance Reviews

A thorough review of the target’s contracts with vendors, suppliers, and customers may become essential to ensure that legal obligations related to data rights, liability, and indemnification are adequately addressed. Appropriate insurance cover for regulatory issues and associated risks will also be crucial.

Since the Intermediary Amendment Rules add new dimensions to due diligence and other compliance requirements, investor decisions related to capital allocation may need to assess data governance maturity. Investors may also need to assess a target’s capacity to design and implement SGI handling, content labeling, takedown and redressal capabilities, as well as proper verification mechanisms, within its core stacks. Companies that can demonstrate scalable compliance with such measures are likely to benefit from reduced regulatory uncertainty and enhanced trust.

Emerging Business Opportunities

While the DPDP Framework introduces compliance challenges for AI ecosystems, it may also spur innovation and create new business opportunities.

Consent Managers

The statutory role of consent managers for managing user consent across platforms may create new business opportunities powered by AI and AI agents.

Privacy Enhancing Technology

The DPDP Framework’s strict compliance requirements may drive demand for AI-driven privacy infrastructure for enterprises. Such infrastructural capabilities may include automated breach notifications, organizational/ internal consent management systems, vendor accountability tools, and retention governance through unified platforms.

Conclusion

Recent regulatory changes in India on data protection and intermediary governance are poised to drive India’s AI ecosystem towards a paradigm based on transparency and traceability. India’s legislative response to AI and data protection issues broadly follows global practices. While certain new trends and regulatory developments appear to raise operational burdens, these are likely to be standard practices for most global organizations. For Indian enterprises, this phase provides an opportunity to plan for, and invest in, appropriate data governance, breach and consent handling, privacy-preserving techniques, as well as SGI verification and labeling systems – including for the purpose of securing future-proofed abilities to navigate evolving data and AI-related requirements in India.

Taken together, India’s DPDP Framework and intermediary regime provide important foundations for AI governance, particularly by imposing discipline around the collection and processing of personal data and by introducing greater transparency, traceability, and accountability for AI-generated or manipulated content. However, their application to ‘agentic’ AI (i.e., systems capable of autonomously planning, initiating, and executing complex sequences of actions – and not merely generating outputs) raises a more fundamental question concerning the adequacy of regulation designed around relatively discrete acts of data processing or content dissemination.^[5]

An AI agent may continuously access, combine, infer from, and act upon information from multiple sources in pursuit of a broader objective, with the particular data used, the sequence of operations, and downstream actions, determined only as such agent operates. This may complicate the application of purpose limitation and data minimization principles, particularly where the agent identifies new information or pursues an unforeseen route to achieve an objective. Recent work on AI purpose limitation in the EU context identifies a regulatory gap between (i) the regulation of data used to train AI, and (ii) the subsequent circulation, reuse, and application of the resulting model. The problem becomes more pronounced with agentic systems because the relevant data processing may evolve dynamically during deployment, rather than being confined to the relatively identifiable processing activity contemplated when the system was designed or deployed.

Agentic AI also complicates the intermediary framework because an agent may simultaneously interact with platforms, generate or transmit content, retrieve information, invoke external tools and services, and take actions on behalf of a user or enterprise. Accordingly, the regulatory characterization of the relevant actor may vary depending on which part of the agent’s activity is examined. An intermediary-oriented framework can address the content that an agent produces or transmits, while the DPDP Framework can govern personal data processed in the course of that activity. However, neither regime may necessarily capture the agent’s autonomous selection of what information to access, which tools to invoke, what inferences to draw, or what sequence of actions to undertake.

Recent scholarship on agentic AI similarly identifies difficulties in applying conventional controller/ fiduciary-processor models where agents exercise discretion over the practical means of processing, while policy analysis increasingly emphasizes lifecycle oversight, meaningful human intervention, and system-level accountability. This creates a potential governance gap between data accountability and behavioral accountability: establishing the fact that data was lawfully processed, or that content was appropriately labelled, may still leave unanswered whether the agent was appropriately authorized to act, whether its actions remained within the intended scope of such authorization, and which actor should bear responsibility when several systems jointly produce the outcome.

These issues suggest that the growing deployment of agentic AI may require India’s existing DPDP and intermediary frameworks to operate as components of a broader lifecycle governance architecture. Such an architecture would need to account for the degree of autonomy granted to an AI agent, the data and systems to which it has access, the purposes for which it may use or reuse information, the external actions it is authorized to undertake, and the ability of its developer or deployer to reconstruct and intervene in its decision-making process. This is particularly relevant as agentic systems move beyond information generation into consequential transactions and operational decision-making. Current developments in India, including work towards enabling AI agents to undertake UPI payments (subject to spending limits, identity controls, and liability provisions), illustrate how quickly these questions may become operational rather than remain theoretical. While the DPDP Framework and the intermediary regime may both remain necessary elements of India’s AI governance ecosystem, their effectiveness in governing agentic AI may ultimately depend on complementary rules addressing autonomy, authorization, traceability, continuous oversight, and allocation of responsibility across the AI value chain.

*Dr. Deborshi Barat is the Head of Public Policy and Knowledge Management at S&R Associates.

[1] For a discussion on ‘intermediaries’ and AI-related advisories under the Intermediary Guidelines, see here, here, and here.

[2] For analyses on deepfakes and synthetic content, see here, here, and here. For a discussion on AI-related opportunities in India in the context of creating ‘synthetic’ content through the use of AI/ ML, see the note here.

[3] See, e.g., here and here.

[4] See here.

[5] For a discussion on opportunities, risks, and evolving legal frameworks related to agentic AI, see the note here.