India AI Impact Summit: A Middle-Power Model for Global Governance

The Delhi AI Impact Summit in February 2026 was the fourth stop in a summit cycle, after the UK’s AI Safety Summit at Bletchley Park in 2023, the AI Seoul Summit in South Korea in 2024, and the AI Action Summit in Paris in 2025. The change in summit titles signals a shift in the priorities of governments and firms. “Safety” centred on risk mitigation and guardrails; “Action” brought geopolitical and geoeconomic competition to the foreground, including the race for standards, markets, and national champions; and “Impact” moved attention to delivery: who can build domestic AI capability for public services and businesses rather than depend on a small number of foreign platforms to do so.

Delhi made that shift unmistakable. This global summit brought together around 20 heads of state or government, delegations from more than 110 countries, and approximately 30 international organisations, alongside chief executives and leaders of organisations building advanced general-purpose AI models. It was also presented as the first-ever global AI summit hosted in the Global South. For foreign policy watchers, the message was clear: AI governance is no longer only shaped by US big tech, China’s state-led approach, and the EU’s regulatory power.

A middle-power proposition, not a spending race  Most governments cannot outspend big tech on chips, cloud capacity, and energy-hungry data centres. Delhi’s proposition was that states can still shape their AI futures through public investment, targeted governance, and partnerships that build domestic capability. The host underlined this with the MANAV framework: Moral, Accountable, National, Accessible, and Valid. In plain terms, “impact” was framed as technology that is ethically grounded, governed, accessible, and legitimate where it is deployed.

Two key takeaways stand out: 1) Inclusion as infrastructure  In Delhi, “democratic AI” was discussed less in the vocabulary of rights and more in the language of access: access to compute, data, models, and multilingual tools so that start-ups and small firms can innovate. The summit favoured compact, cost-efficient systems designed to run under real constraints. Examples often cited included Adalat AI for court transcription and Bhashini, a public translation platform aimed at making digital services usable across linguistic diversity.

2) Control over key models  Delhi also brought into focus the shift from “AI as applications” to “AI as capability.” Governments are uneasy about building essential services on external foundation models where licensing terms, export controls, or abrupt product changes can become a pressure point. The summit highlighted investment in domestic model capacity as an option for middle-income states. Sarvam AI is emblematic, with an important caveat: Sarvam-M is built by post-training a foreign open-base model, specifically Mistral, rather than training a base model from scratch. Even so, the ambition is to use public support, dedicated compute, and local language data so that key models can be audited, adapted, and maintained domestically, reducing long-term dependence on a small number of providers.

What Delhi signalled for Germany   The Delhi Summit marked a pivot point for the Germany-India digital partnership by setting out a pragmatic “AI Pact.” Since 2017, the bilateral digital dialogue has been broad, and the new agreement is designed to translate it into concrete projects anchored in the physical economy: mobility, energy, and healthcare. For Germany, the interest is twofold. First, it seeks to connect its SMEs with India’s developer ecosystem to solve “frugal AI” challenges, meaning systems that are high performance but low compute. Second, the pact provides a bridge to the EU AI Act by keeping European regulatory standards interoperable with India’s emerging guidelines. This reduces the risk of a “regulatory island” effect and allows for protection without protectionism. By focusing on shared computing capacity and specialised language models, Europe is signalling that its view of AI is not limited to a US–China lens – rather, it envisions AI as a collaborative infrastructure project with middle-power partners.

The missing centre: Civil society, climate, and jobs  The Delhi Declaration, as reported, reflects a hybrid approach. It argues for innovation, but it also calls for accountability, sustainability, and transparency that people can understand. Yet civil society voices remained less prominent than those of governments and industry. That matters for democratic accountability in the digital state: citizens need credible protections and remedies vis-à-vis state overreach and unchecked AI deployment by private actors. Two unavoidable issues underscore this gap, the first being energy: data centres used roughly 1.5 per cent of global electricity in 2024, and demand is expected to rise sharply by 2030 as AI workloads grow. The second is jobs: major employer surveys suggest that 23 per cent of jobs will change by 2027, with significant churn even where net losses are smaller.

The most prominent voices at the summit in Delhi spoke often about skills and opportunity, but less about worker protection, clear routes for complaints and remedies, and accountability when systems fail or exclude people at scale. If “impact” is to mean more than rollout, future summits should place civil society at the centre alongside government and industry, so energy use, worker security, and public accountability stay on the main agenda.