The India AI Impact Summit closed today with 75 nations signing the Delhi Declaration, a non-binding agreement affirming that “AI’s promise is best realised only when its benefits are shared by humanity.” Hours earlier, White House official Michael Kratsios told the same summit: “We totally reject global governance of AI.”
That contradiction - a sweeping declaration of shared purpose sitting next to an explicit American refusal to be governed by it - captures exactly where international AI policy stands in February 2026. Everyone agrees AI should benefit humanity. No one agrees on who gets to make the rules.
What the Delhi Declaration Actually Says
Union IT Minister Ashwini Vaishnaw confirmed that “all major AI-relevant countries” signed the declaration, which the European Union endorsed through Executive Vice-President Henna Virkkunen.
The declaration’s three priorities:
Population-Scale AI: Building systems that work in local languages and serve diverse populations, not just English-speaking users in wealthy countries. India backed this with BharatGen Param2, a 17-billion-parameter model supporting 22 Indian languages released during the summit.
Data Sovereignty: Ensuring countries retain control over their own data, with calls for building domestic “Sovereign AI” capabilities linked to national Digital Public Infrastructure.
Green AI: Promoting energy-efficient computing infrastructure and shared climate-modeling data.
Each priority is reasonable. None includes an enforcement mechanism. The declaration is a statement of principles, not a treaty - which is exactly why 75 countries could sign it, and exactly why the US was comfortable not blocking it. Non-binding declarations don’t threaten anyone’s sovereignty or business model.
The US Position: No Rules, Just Exports
The American delegation didn’t come to New Delhi to negotiate governance frameworks. It came to sell.
Kratsios’s “totally reject global governance” statement was paired with promotion of US AI exports and the concept of “strategic autonomy alongside rapid AI adoption.” Translation: buy American AI technology and build your own capacity with it, but don’t try to regulate what American companies can do.
That message found its most concrete expression in Pax Silica, a US State Department initiative that India formally joined on the summit’s final day. IT Minister Vaishnaw and US Under Secretary of State for Economic Affairs Jacob Helberg signed the agreement, adding India to a coalition that includes Australia, Greece, Israel, Japan, Qatar, South Korea, Singapore, the UAE, and the UK.
Pax Silica covers the full technology stack - raw materials, power generation, chip fabrication, data centers, frontier AI models, and global logistics. The name is deliberate: silicon as the foundation of a new world order, with the US as architect. The framework’s stated purpose is building “resilient supply chains” for critical minerals and AI technology. Its unstated purpose is reducing dependence on China, which controls roughly 90% of global rare-earth processing and 70% of mining output.
India gets tangible benefits. Vaishnaw said Pax Silica “would greatly benefit India’s electronics and semiconductor industry,” noting ten semiconductor plants already established or under construction in the country, with the first commercial plant beginning production soon. Indian engineers are designing 2-nanometer chips domestically.
But joining a US-led supply chain bloc while signing a multilateral declaration on “shared” AI governance is a balancing act. India wants access to American semiconductor technology and investment. It also wants to position itself as the Global South’s representative at the table. Those goals don’t always point in the same direction.
China’s Calculated Absence
China, the world’s second-largest AI power, was almost entirely absent from the summit. Officials attributed this to the overlap with Chinese New Year. But the timing of a five-day summit that conveniently excluded Beijing from the room while India signed onto a US-led anti-China supply chain initiative seems unlikely to be coincidental.
China’s absence simplified things for everyone else. It let India sign Pax Silica without an awkward scene. It let the US promote its vision without direct pushback from its primary competitor. And it let the Delhi Declaration pass without the friction that Chinese participation would have introduced.
The result is a set of summit outcomes that look multilateral on paper but are shaped by a bilateral dynamic: the US and India strengthening their technology partnership while China watches from the sidelines.
The $3 Billion That Hasn’t Materialized
UN Secretary-General Antonio Guterres offered the summit’s sharpest critique and most ambitious proposal. He called for a $3 billion global fund to ensure developing nations aren’t “logged out” of the AI age, and delivered a pointed message to the tech executives in the room.
“The future of artificial intelligence cannot be decided by a handful of countries or left to the whims of a few billionaires,” Guterres said. His proposed $3 billion target, he noted, is “less than one per cent of the annual revenue of a single tech company.”
The fund would focus on skills development, data capacity, affordable computing power, and inclusive ecosystems for countries that currently lack AI infrastructure. It’s a sensible proposal. It’s also unfunded. No country or company made a binding commitment to the initiative during the summit.
Compare Guterres’s ask with the private sector pledges. Adani committed $100 billion. Reliance pledged $110 billion. Microsoft offered $50 billion for the Global South. But those are corporate investments with expected returns, flowing to commercial infrastructure in countries with existing markets. Guterres was asking for $3 billion in what amounts to aid - and got applause instead of checks.
Middle Powers Running Scared
Beneath the handshakes and declarations, a more anxious conversation played out among what diplomats call “middle powers” - Europe, Canada, and India itself.
These countries are increasingly alarmed by their dependence on American AI technology. The Trump administration’s recent actions regarding Greenland and NATO have heightened the sense that US-allied nations can’t take Washington’s goodwill for granted. European delegates at the summit spoke about developing independent AI capabilities rather than relying on Silicon Valley.
French President Emmanuel Macron arrived in New Delhi calling for Europe to “shape the rules” of AI with allies. But shaping rules requires either regulatory power that can reach across borders - which the US has rejected - or domestic AI capabilities that can compete with American and Chinese systems. Europe has neither at the scale required.
India’s position is stronger. It has domestic capital, engineering talent, and now semiconductor manufacturing ambitions backed by Pax Silica membership. But it also has 1.4 billion people who need AI systems that work in their languages, a government navigating between Western technology partnerships and non-aligned instincts, and an opposition that’s calling the summit a “disorganized PR spectacle.”
Civil Society Shut Out
One pattern went largely unremarked in the summit coverage: the structure granted multinational corporations parity with sovereign governments through the CEO Roundtable and Leaders’ Plenary, while providing no equivalent high-level platform for civil society, labor representatives, or human rights organizations.
The people most affected by AI deployment - workers whose jobs are being automated, communities subjected to algorithmic surveillance, patients relying on AI medical systems - were absent from the rooms where commitments were made. Tech executives sat alongside heads of state. Labor leaders and digital rights advocates did not.
This isn’t unique to the India summit. It’s the pattern of every major AI governance event. But it matters more when the stated goal is ensuring AI benefits “all of humanity.” You can’t credibly claim to represent everyone while excluding most of them from the conversation.
What the Summit Actually Produced
Strip away the speeches and photo opportunities, and the India AI Impact Summit delivered three concrete things:
A declaration without teeth. The Delhi Declaration articulates principles that 75 nations can agree on because agreeing costs nothing. No enforcement, no penalties, no independent verification.
A supply chain alliance with real consequences. India’s entry into Pax Silica reshapes semiconductor and AI hardware supply chains in ways that will affect chip availability, pricing, and technology access for years. This is the summit’s most consequential outcome, and it happened in a bilateral signing, not the multilateral main event.
A proof of concept for Global South summitry. India demonstrated it can convene a global AI event and extract hundreds of billions in investment commitments. Whether those commitments materialize is another question, but the convening power itself has value.
Where This Leaves AI Governance
The honest answer: fragmented. The US won’t accept international regulation. China wasn’t in the room. Europe wants rules but lacks the technology to enforce them. India wants to be the bridge but just joined a US-led bloc. And 75 nations signed a declaration that asks for “shared benefits” without defining how to share them or what happens when someone doesn’t.
The gap between what the Delhi Declaration aspires to and what Pax Silica actually does tells the real story of AI governance in 2026. The multilateral track produces aspirational language. The bilateral track produces binding partnerships. The countries that matter most are choosing the bilateral track while paying lip service to the multilateral one.
That’s not necessarily cynical - it’s how geopolitics has always worked. But it means the people placing hope in international AI governance frameworks should look carefully at what happened in New Delhi this week. Seventy-five signatures on a declaration that means what each signatory wants it to mean. A $3 billion fund proposal with zero committed dollars. And one superpower telling everyone, openly, that it rejects the premise.
The AI world order isn’t being built at summits. It’s being built in bilateral deals, semiconductor factories, and the supply chain agreements signed in side rooms while the main stage produces declarations that read well and bind no one.