As India solidifies its position in the global AI race, a critical debate is taking shape. The government is now considering a major policy shift: subjecting advanced AI models to safety checks before they can be launched to the public.
A New 'Safety First' Framework
The Indian government
is exploring a significant pivot in its approach to artificial intelligence, moving from a light-touch advisory role to considering a formal legal framework. At the heart of this discussion is a proposal to create a separate, more stringent safety regime for what are known as 'frontier' AI models. These are the most powerful and capable systems, like those that power advanced chatbots or generate highly realistic synthetic content, which officials believe pose risks that warrant new legislation. The talks, confirmed by Union Electronics and IT Minister Ashwini Vaishnaw, are part of preparations for a forthcoming consultation paper that will seek input from industry and the public before any law is drafted. This signals a belief within the government that the risks associated with AI have crossed a threshold requiring more direct oversight.
What Would Pre-Launch Checks Involve?
The proposed 'checks' could involve several layers of scrutiny for developers of frontier AI models. Among the measures being examined are pre-release evaluations to assess a model's capabilities and potential for harm before it reaches the public. This could include mandatory testing in controlled 'sandbox' environments, where regulators and researchers can study a model's behaviour and identify vulnerabilities in a lab setting rather than in the wild. Officials are also discussing security assessments proportionate to a model's power and potential restrictions on certain functions or users, depending on the level of risk identified. The central idea is to implement a vetting process that allows for the discovery of serious issues—like the potential to enable cyberattacks or create convincing deepfakes—before a model is widely deployed.
The Global Context for AI Regulation
India's potential move does not exist in a vacuum. Governments worldwide are grappling with how to regulate AI, creating a complex and diverging global landscape in 2026. The European Union is already implementing its landmark EU AI Act, a comprehensive law that categorizes AI systems by risk and imposes strict obligations on high-risk applications, such as those used in employment or for biometric identification. The United States has seen a mix of federal guidance and a growing number of state-level laws, like those in California and Colorado, which enforce transparency and risk assessments. China, meanwhile, has also enacted targeted AI laws. By considering pre-deployment checks, India is signalling its intent to join the club of nations taking a proactive, safety-conscious stance, rather than relying purely on self-regulation by tech companies.
Balancing Innovation and Responsibility
The proposal has ignited a crucial debate within India's tech ecosystem. On one hand, there is a clear push for safety, driven by rising concerns about AI-assisted cybercrime, misinformation, and other societal harms. India's Computer Emergency Response Team (CERT_In) has issued high_severity warnings about the risks of frontier AI models. On the other hand, there are concerns that overly burdensome regulation could stifle innovation and place Indian startups at a disadvantage. The challenge for policymakers will be to define what qualifies as a 'frontier model' and design regulations that are robust enough to ensure safety without choking the growth of a strategically vital industry. The distinction would allow regulators to focus intense scrutiny only on the models capable of causing significant harm, rather than burdening every AI developer with the same compliance requirements.
