India’s rapidly growing technology sector is facing a potential regulatory shift. A new legal framework is being considered that would place developers of the most advanced artificial intelligence under a microscope, signalling a major policy evolution.
What is Frontier AI?
The term 'frontier AI' refers to the most powerful and advanced AI models at any given time. These are not your everyday chatbots or recommendation algorithms. Instead, think of large, general-purpose systems like GPT-4 and its successors, which demonstrate emergent capabilities in complex reasoning, coding, and understanding across various types of information. Because these models push the boundaries of what AI can do, they also introduce new and significant risks, including enabling sophisticated cyberattacks or generating highly convincing fake content, often called deepfakes. This has prompted governments worldwide, including in the EU and US, to consider specific oversight for this class of technology.
A Shift Towards Risk-Based Regulation
India is reportedly moving from a light-touch approach to a more structured, risk-based legal framework. Officials have indicated that the government is preparing a consultation paper on AI regulation that will likely form the basis of a new law. The core idea is to classify AI systems based on the level of risk they pose. Low-risk applications, such as basic productivity tools, would face minimal obligations. However, 'high-risk' systems, especially those deployed in critical sectors like healthcare, finance, and banking, would be subject to much stricter rules. This tiered approach aims to avoid stifling innovation in less sensitive areas while focusing regulatory energy where potential harm is greatest.
What 'Extra Scrutiny' Could Mean
For developers of frontier AI, this 'extra scrutiny' could involve several new requirements before their models are released to the public. Measures being examined include mandatory pre-release evaluations to assess a model's capabilities and potential for misuse. Developers might also be required to test their systems in controlled 'sandbox' environments to identify and mitigate risks before a wider rollout. Security assessments proportionate to a model's power are also on the table. The goal is to ensure that developers have builtin appropriate safeguards rather than reacting to problems after a model is already in public use.
The Rationale: Balancing Innovation and Safety
The push for a dedicated AI law reflects a growing belief among Indian policymakers that the risks associated with AI have crossed a significant threshold. Union IT Minister Ashwini Vaishnaw has confirmed the government's intent to create dedicated legal guardrails, particularly to tackle threats from deepfakes, automated misinformation, and the unpredictability of advanced AI systems. Concerns about national security and cyberattacks are also driving this conversation. India's Computer Emergency Response Team (CERT-In) has issued warnings about how frontier models could be used to discover software vulnerabilities and automate phishing attacks. The challenge for lawmakers is to craft a framework that addresses these serious risks without hampering India's ambition to be a global leader in AI development.
Global Context and What's Next
India's proposed approach aligns with a global trend toward more direct oversight of powerful AI. The European Union's AI Act, for instance, also uses a risk-based classification and imposes specific obligations on developers of high-impact models. In the past, recommendations from bodies like the Telecom Regulatory Authority of India (TRAI) suggested creating a dedicated statutory body, the 'Artificial Intelligence and Data Authority of India' (AIDAI), to oversee the sector. While a draft law is yet to be written, the upcoming consultation paper will be a critical first step. It will seek feedback from the industry, civil society, and other stakeholders on defining 'frontier models', the scope of pre release testing, and how to assign liability when AI systems cause harm.
