India’s DPDP Act & copyright law can attract AI data centres

By Pune Media On Apr 16, 2025

Meta is hoping that the exemption for scraping of publicly available personal data under the Digital Personal Data Protection Act, 2023 (DPDP Act), and provisions under India’s Copyright Act would make the country a friendly jurisdiction to train AI models, the company’s Public Policy Director for Data Governance and Emerging Technologies, Sunil Abraham said at the Global Technology Summit on April 12.

For context, India’s data protection law does not apply to publicly available personal data. Abraham explained that the exemption applies when someone discloses this personal data by themselves. He argued that perhaps the intent to exempt publicly available personal data from the scope of the DPDP Act was to allow for model training. However, in practice, it can be difficult for companies to verify that a user has made their data publicly available by themselves. “A small modification [to the data protection framework could be one] which allows the data fiduciary to test whether the publicly available data has been made legally available,” Abraham said.

He explained that if the Government made this modification, some people believe that Section 52(1)(c) of the Copyright Act and this exemption in the DPDP Act would make India “perhaps the most friendly jurisdiction on the whole planet for you to bring your data centres and train your models. So that’s roughly where we’re hoping we go, if that’s possible,” Abraham added.

What is Section 52 of the Copyright Act?

Section 52 of the Copyright Act lists the activities that the law does not consider copyright infringement. Part 1, sub-section (c) of this section says that “transient or incidental storage of a work or performance for the purpose of providing electronic links, access or integration” are not copyright infringement in cases where the copyright holder does not expressly prohibit such links, access or integration. Responding to Abraham’s comment, Rahul Matthan, a partner at Trilegal, said that this clarification is important from the perspective of leveraging AI.

How does Section 52 apply to AI?

“I personally believe this argument [of Section 52(1)(c)-based exemption] for AI data centres from copyright infringement would emerge if the primary function that the storage is performing is not infringing then incidental storage is also not infringement. This means that if you are storing data to train an AI model and the objective act of the training is transformative, then arguably, such storage is also exempt from copyright protection,” Sneha Jain, Partner at Saikrishna Associates, explained.

However, she mentioned that the subject of copyright and AI was currently under examination as part of the OpenAI vs ANI case in the Delhi High Court (HC). In an amicus curiae submission before the court on March 10, lawyer Adarsh Ramanujan said that India’s copyright law does not have a text and data mining exemption (TDM). Ramanujan had also mentioned that the fact that something is freely available does not take away copyright protections. Splitting the functioning on an LLM into three parts—data collection, tokenising collected data and training the model— he argued that step one and three could both amount to infringement.

Jain mentioned that it is necessary to reconcile the DPDP Act with the copyright law because while the latter protects copyrighted works, like people’s personal data compiled into a list, the former allows for the processing of such data if it is publicly available.

Other key points from the discussion:

Expect DPDP Rules in six to eight weeks:

Bhuvnesh Kumar, the Additional Secretary for the Ministry of Electronics and Information Technology (MeitY), mentioned that the Digital Personal Data Protection Rules (DPDP Rules) — which operationalise the Act— are largely ready, and the legislative process and vetting of the rules should be done in six to eight weeks. He further mentioned that the Government received 6,000 comments during the consultation on the rules. Besides the comments that the Government got on the MyGov portal, it also received about 100 written representations.

Dealing with data minimisation as a social media platform:

Speaking to Abraham, Matthan mentioned how a platform like Facebook has a repository of a person’s whole life through, for example, their photographs. He asked Abraham how a platform like Facebook will deal with the requirement of data minimisation as envisaged under the rules. For context, the DPDP Rules state that people can ask service providers to delete their data. Further, the rules state that social media platforms have to delete a user’s personal data if said user has not engaged with their service for three years.

Data deletion could cause IP loss:

Abraham argued that deletion using an algorithmic software could cause a risk to copyright. This, in turn, could make it a violation of the DPDP Act/Rules. Matthan, however, mentioned that platforms are not obliged to hold user data and that users could store it themselves.

“They have transferred it to me [Meta] on a license, and I’m a licensee of their IP [intellectual property]. It might be the only version of that IP they own. The terms of service that they have with me is an expectation that I’m not going to randomly delete their intellectual property,” Abraham explained.

For You

[ad_1]

_{Images are for reference only.Images and contents gathered automatic from google or 3rd party sources.All rights on the images and contents are with their legal original owners.}

_{Aggregated From –}

_{Source link}

[ad_2]