A key provision in India’s Digital Personal Data Protection (DPDP) Act, 2023, could make it legally risky and in some cases unfeasible for Indian companies and startups to train artificial intelligence models, the Internet and Mobile Association of India (IAMAI) has warned in a new submission to the IT ministry.
In the submission, a copy of which Moneycontrol has reviewed, the industry group has pointed to Section 3(c)(ii) of the law, which says that personal data made “publicly available” by an user or under legal obligation is outside the scope of the Act.
The IAMAI argued that the clause is too ambiguous for developers working with large datasets that may include scraped or publicly posted information.
"It can be challenging to ascertain whether a Data Principal (user) has voluntarily made their personal data public. For example, platforms often provide privacy settings that allow users to modify the visibility of their data. A Data Principal might have initially shared their data publicly and later restricted its visibility, complicating the determination of whether the data was voluntarily made public," IAMAI said.
Companies which develop LLMs usually train their models on massive amounts of data often from open platforms or government sources. Any ambiguity in legal provisions can lead to "undue compliance burdens, hinder technological progress and ultimately obstruct the realisation of AI's potential," IAMAI argued.
The trade body which represents a swathe of digital companies both Indian and global, including members of Big Tech firms, recommended the government to either amend the clause and clarify AI model training is exempted from its provisions, or to Section 17(5) of the Act to issue an interim exemption for model training and fine tuning.
"We acknowledge that amending the Act may require time. As an interim measure, we recommend that, in exercise of the powers conferred under Section 17(5) of the DPDP Act, the Central Government exempt data fiduciaries from the Act’s provisions, provided the data processing is solely for the training or fine-tuning of AI models and is limited to publicly available personal data," the IAMAI said.
IAMAI warned that India may fall behind in the global AI race, at a time when countries like the US and China had lower compliance hurdles for AI developers.
"Given that other jurisdictions such as the United States and China have made a concerted effort to minimise regulatory burden on AI businesses in the form of data protection, the Indian AI ecosystem may find itself to be less competitive. Therefore, it is imperative to minimise the burden on the ecosystem, to give it the conditions to leapfrog the global leaders in AI, like Indian startups have done in other sectors as well," it added.
The submission reflected the view of most IAMAI members, though it noted that Reliance Jio and Culver Max Entertainment have expressed different opinions.
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
