Can US AI copyright law impact Indian policymaking?

The proposed Act suggests compensating original creators for content that is used to train Gen Al. This can be a good middle ground for many in tech to follow

- VERNIKA AWAL

On 9 April, Adam Schiff, a Democrat Senator of the United States of America, introduced a new Bill in the US Senate. Titled Generative AI Copyright Disclosure Act, 2024, the Bill sought to raise into mainstream legal discourse conversations around the usage of copyrighted data in training generative AI algorithms. In turn, what this draft legislation has done is underline the need of the hour to make generative artificial intelligence (Gen AI) a commercially viable business—even in terms of how it works in India, and the country’s proposed Digital India Act (DIA).

WHAT DOES THE LAW SAY?

The first and foremost proposal of the Bill is to make Big Tech firms such as Google, OpenAl, Meta and Microsoft accountable for the data that they consume and subsequently put to use for commercial gains. To do this, the Bill suggests that companies developing large, foundational AI models must disclose the datasets on which these models are trained.

This is crucial because right now, most Big Tech firms have offered vague, obscure definitions for where most of their data is sourced from. In various aspects, the source of the datasets used in generative AI have been referred to as ‘black boxes’—where the data cannot be traced, or explained in clear terms to individual and enterprise users alike.

To clarify this, the Bill has proposed setting up a centralised ‘register’, which will be maintained by the central US government. Companies headquartered in the US, building any AI model, will be required to disclose their sources of data to this register, at least 30 days before the model is published in the open.