The Indian government is considering whether to set up an Artificial Intelligence and Safety Institute (AISI) that can help set standards, frameworks and guidelines for AI development without acting as a regulator or stifling innovation. A senior government official told those involved in the talks. At least seven people with knowledge of the matter spoke to HT on condition of anonymity at the October 7 meeting.
The consultation, led by Abhishek Singh, Additional Secretary, Ministry of Electronics and Information Technology (Meity), which oversees Indian AI, sought opinions on how the Indian AISI should be structured, what its mandate should be, and its mandate. It was a preliminary meeting to seek the How can we collaborate with other AISIs around the world?
The UK announced an initial investment of £100 million (~ INR110 billion). It is closely followed by the United States, which was established as part of the National Institute of Standards and Technology (NIST). Japan launched AISI in February 2024. The UK AISI is within the government and has a mandatory element, while the US AISI is primarily a standards-setting body.
In May 2024, the European Union and 11 countries, including the United States, United Kingdom, South Korea, Australia, Canada, France, Germany, Italy, Japan, South Korea, and Singapore, signed the Seoul Declaration at the Seoul Summit, which states: Among other things, it aims to create or expand AISI, research programs and other related institutions to foster collaboration in safety research and share best practices.
Ahead of the Oct. 7 conference, MeitY shared two categories of questions with stakeholders, including companies such as Meta, Google, Microsoft, IBM, and OpenAI. Industry associations such as Nasscom, Broadband India Forum, and BSA-The Software Alliance. Multiple IITs. Consulting companies such as The Quantum Hub and Dialogue. It also includes civil society organizations such as the Digital Empowerment Foundation and Access Now.
The first category of questions covers AISI’s focus, its core purpose, the organizational structure that best suits its mission and scalability, and how it can develop its own AI safety tools “tailored to India’s unique challenges.” And it was about who was strategically responsible for AISI. partner.
Second, how AISI can “build strong partnerships and gain stakeholder support,” i.e., what strategies to engage key stakeholders to support AI safety. We focused on how AISI can establish and maintain effective domestic and international partnerships, and what role it should play in global AI. Safety discussions and standards.
HT’s conversations with participants first suggested that there was a consensus that the government was serious about establishing an AISI. Second, AISI is not about regulation; it is about identifying hazards, identifying risks, and setting standards, which can ultimately influence future regulation. Third, we need interoperable systems to prevent silos.
AISI also has the potential to advance risk assessment toolkits and voluntary compliance toolkits available to industry. The framework is not binding as the Government does not intend to establish AISI as a regulatory body, but this is subject to change depending on the input the Government receives from stakeholders. AISI-supported research could be used to inform the eventual AI policy of the Indian government.
Experts aware of the government’s thinking on AISI explained to HT that three questions need to be asked to underpin India’s AISI debate. First, why does India need AISI? Second, how should it be configured? And third, how does India want to view AI security? In other words, what use cases and applications should India’s AISI consider high-risk or low-risk? Should we focus on risk, innovation, or a combination of the two?
This person said that the main purpose of AISIs established around the world is to understand and reduce risks associated with the development and deployment of AI, and to mutually share parameters for quantifying risks. I explained.
structure
At a meeting on October 7, stakeholders discussed how to structure the Indian AISI and whether the AISI, or at least its secretariat, should be located within Meity. Officials said AISI should have reasonable independence and could therefore be set up within academic institutions such as IITs.
It was also discussed what role states can play, especially as several states, such as Telangana, have started using AI for public service delivery. AISI could provide guidance (not prescriptive) and best standards on AI use, the officials said.
budget
At the October 7th meeting, there was a brief discussion on budget allocation for the Indian AISI. Shin, it is understood, said about INR$20 billion has been allocated for the safe and reliable AI pillar of the IndiaAI mission. This could be used for AISI and if that is not enough, more funds could be recycled from other arms of the IndiaAI mission.
In March, the Cabinet approved the following budget expenditures: INR10,372 crores for IndiaAI Mission over 5 years. In the fiscal year 2025 budget announced in July, the government allocated the following budget: INR551.75 million for IndiaAI Mission.
delegation
Initially, at least one official said at the meeting that AISI’s mandate could be limited to setting standards for AI, understanding the risks of AI development and deployment, and how to conduct international cooperation. Initially, not limiting the agenda could be overwhelming for AISI, officials said.
