One major point of discussion, a source said, was the liability framework for companies developing frontier AI models and tools, and whether safe harbour protection should be extended to them.
“We are looking at what the contours need to be. The general understanding is that today, all of these companies fall under the broad omnibus category of internet intermediaries. The argument that because they just provide the tool and, therefore, are not responsible for the content being generated by using that platform or tool is not a fair one,” an official said.
Among other key points being discussed for the proposed AI law are broader provisions for regulating synthetic media, a consent mechanism for data used by frontier AI tools and models, and ownership of data processed by these large and small language models (LLMS and SLMs), the sources said.
Earlier this year, the government formed a 10-member AI Governance and Economic Group (AIGEG) chaired by Union Minister for Electronics and Information Technology (Meity) Ashwini Vaishnaw, as well as a sub-committee under AIGEG called the Technology and Policy Expert Committee (TPEC).
The TPEC, chaired by Meity Secretary S Krishnan, counts B Ravindran of the Indian Institute of Technology (IIT), Madras, Rajat Moona, the director of IIT, Gandhinagar, representatives from industry bodies such as Nasscom, the Data Security Council of India (DSCI), and the Manufacturers’ Association for Information Technology (MAIT) as members.
The discussions on the new legal framework for AI are being held by both the AIGEG and the TPEC.
Formed in April this year, the AIGEG is also expected to sharpen India’s broader AI strategy by funneling the bulk of funding into a limited number of use cases capable of delivering measurable outcomes within the next 12-18 months. Priority sectors are likely to include health care, agriculture and education.
The need for an apex AI governance body in India follows growing concern among experts over the rapid advancement of LLMs and tools such as Claude’s Mythos, which have demonstrated the ability to penetrate legacy software systems and existing digital defences.
Both AIGEG and TPEC have suggested that the AI regulatory framework should initially focus on “high-impact sectors such as finance and public services”, another official said. This comes against the backdrop of reports about the capability of frontier AI models such as Mythos to find loopholes and bugs in legacy banking software, and the accompanying risks.
To accelerate India’s access to such frontier models like Mythos and OpenAI’s ChatGPT 5.6, the committees have suggested expediting the set up of the AI Safety Institute (AISI), the official said.
“The idea is that the AISI, which will be government-controlled with stakeholders from all areas, will be better equipped to coordinate and deliberate on the finer nuances of access to these frontier models,” said one of the officials quoted above.
Apart from bringing provisions to answer these questions, the government may also prescribe a broad AI risk framework that must be followed by companies that are looking to launch new AI-enabled services in the country.
It would define the broad contours as well as dos and don’ts for AI tools and services, besides establishing the sandbox within which such products should be developed, said a senior government official.
-
Government eyes regulatory gap analysis of the IT Act, 200 and AI governance framework
-
Discussions also underway on regulations for synthetic media, AI model data ownership
-
AIGEG, TPEC suggest establishing AI Safety Institute soon
-
AI regulatory framework likely to focus on finance and public services
 "AI, artificial intelligence")