On the ground at India’s AI Impact Summit – part 1

The New Delhi summit was a game of two halves

India’s AI Impact Summit in New Delhi was a game of two halves, writes OpenUK’s Amanda Brock.

As I landed in India I was greeted at the airport by signage welcoming open source to the AI Impact Summit, from OpenUK and its local partners.

Last week’s AI impact Summit began with an inclusive first few days. As the first Summit held in the Global South, India wanted to make a splash with the event, held in New Delhi at the Bharat Mandapam campus.

With rumours of 150,000 attendees and the total number certainly being tens of thousands, the Summit was a massive undertaking. Planning began in August 2025, with the Summit organisers announcing a summit application process for content both before and during the Summit.

A little chaotic perhaps in its organisation, the Impact Summit’s laudable Sutras - people, purpose and planet - were emblazoned across the site, and the first few days seemed to be accessible to all.

India’s culture is hard to ignore - extreme wealth to extreme poverty in one glance. All were represented in the early stage of the Summit, with busloads of women having been brought to the event from rural villages. The women’s queues at security checks were long for these days, and all of society certainly seemed to be represented.

The selected content was also representative. Organisations brought in groups of speakers to participate in panels and workshops. There was a very distinct split in focus between sessions led by local organisations driving the needs of the Global South and the international hosts.

I will challenge again that whilst there are some clear needs to be met for the Global South, such as a lack of infrastructure and needs generally being at a greater scale, many of the perceived challenges - such as language, culture and access to compute - are also being faced by the Global North. We are often not as far apart as perceptions presume.

Localising AI to India's many languages and the use of small language models on phones, over the LLMs of the Global North, makes sense for India. With a population glued to phones, enabling reach via their mobile devices is clearly the best way to reach the stated goal of “access for all.” It was no surprise that the Government-backed Sarvam releases were Small as opposed to Large Language Models.

However, it was more of a surprise that these models were described as open source when they are shared on proprietary licensing with a commitment to make them open weights (not open source) in the future.

OpenUK led a panel and workshop in the Summit itself and also hosted India’s open source leaders from organisations like FOSSUnited and Software Freedom Law Centre, India in a roundtable. Prof Anil Madhavapeddy hosted a hackathon for geospatial AI for good so popular it was followed up the next day with a lecture at IIT Delhi.

In the OpenUK panel on resilience, Anastasia Stasenko, one of President Macron’s 2025 Summit Ambassadors and co-founder of Pleias, broke the concept of ‘access for all enabled by open source’ into two: first, empowerment enabling innovators and entrepreneurs to build on the open innovation; and second, engagement enabling everyone to engage in AI outputs.

Laura Gilbert of the Tony Blair Institute also raised the current challenges of inclusivity, and the shift from safety in the first summit at Bletchley to security today. Jimmy Wales, Founder at Wikimedia, raised the challenges of infrastructure and vehemently agreed with Mishi Choudary, SVP and general counsel at Virtru, on the absolute need for openness and collaboration to enable resilience.

The final theme for discussion was the level of sovereignty needed to empower nation states without closing down the global collaboration at the heart of our present and our future innovation.

When it came to open source and AI openness, OpenUK had kicked off activities last August with a series of events across India, with good understanding and engagement across the open source communities.

At the Summit itself we saw hundreds of people discussing open source, inspired by our work and local collaboration over the last six months. Policy organisations in particular, armed with the information we had provided, led events on openness, digital public good and access. The rooms discussing these topics were full and almost dangerously overflowing.

The focus on healthcare, education and making people’s lives better was real and there was a recognition that this can only be achieved by AI being open.