India's Ministry of Electronics and Information Technology (MeitY) is working on drafting an AI-specific law that would not provide for criminal penalties for violations, as it believes the benefits of AI outweigh the drawbacks, reports The Indian Express.
The law, which could be promulgated as a standalone bill, could mandate social media platforms such as Facebook, Instagram, YouTube and X to add watermarks or labels to content generated using AI.
The ministry is also considering a legal framework that could require companies building large-scale language models to train their databases on Indian languages and “India-specific” content so that the models can better understand the “Indian context,” the official said.
The investigation comes after some AI systems, including Google's Gemini, were found to have given different answers to similar questions about world leaders and Prime Minister Narendra Modi last year. The IT ministry did not respond to requests for comment.
In November last year, Union IT Minister Ashwini Vaishnav had called deepfakes a “new threat to democracy” and said the government would bring in regulations to curb their spread on social media platforms.
Work on the law was halted during elections earlier this year, but it now appears to be gaining new momentum within the ministry. At the time, Vaishnaw said the government's plan had four main pillars: detecting deepfakes, preventing them by removing or reducing their virality, strengthening reporting mechanisms, and spreading awareness about the technology.
In May, IT Secretary S. Krishnan said at an industry event that the government would try to regulate AI, but not at the expense of innovation, suggesting the ministry is expected to take a middle path in regulating AI. The ministry has incorporated some of the more recent tech laws, such as data protection, and has focused on protecting civil rights as well as promoting the country's growing startup base.
“As we try to regulate AI, we are clear that innovation must not be stifled in the process. Innovation needs to be encouraged… As was the case with the DPDP (Digital Personal Data Protection) Act, we will ensure that in future, both the interests of innovation and the protection of vital interests are taken into account,” Krishnan added.
Indeed, some social media and AI companies have said they have begun watermarking AI-generated content on their platforms even before regulations require it. Adobe, Arm, Intel, Microsoft, and Truepic have also launched a collaborative project called the Coalition for Content Provenance and Authenticity (C2PA), which focuses on systems that provide context and history for digital media.
Meta says it is building tools that can identify “invisible markers” in content based on standards set by the C2PA and label images from Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock.
© Indian Express Ltd.
First uploaded: July 17, 2024 4:40 AM
