JACOB HELBERG, the US Under Secretary of State for Economic Affairs and the architect of the Pax Silica initiative, published a manifesto on June 23 attacking the United Nations' vision of "digital sovereignty". Countries that attempt to build their own AI capabilities, he wrote, will achieve only "a kind of synchronized mediocrity - a planet of subscale clones, each heroically reconstructing last year's breakthrough while the breakthrough itself moves on without them." And then the boast: "While others rebuild the present, American firms will be inventing the future."

Read alongside the Pax Silica Summit convened by the State Department in Washington the same week, where the initiative expanded to 24 participating countries and 35 nations co-signed a Joint Statement on AI Opportunity, the essay is more than an opinion piece on AI governance. It is the clearest articulation yet of how the US intends to organise the global technological future: not through national self-sufficiency, but through what Washington calls "trusted interdependence". Helberg repeated the message days later at the IX Leadership Summit of the US-India Strategic Partnership Forum (USISPF) in Washington, warning that the idea of sovereignty was being "weaponised by a number of different political voices overseas" and dismissing as "incredibly backward" the notion that a country is not sovereign unless it controls its own AI stack.

The question every country in the Global South must ask is simple: who controls, who benefits, and who pays? The answers reveal that "trusted interdependence" is dependence organised under imperial command, and that the "trust" flows in only one direction.

First, consider what the doctrine actually proposes. Helberg calls his alternative "innovation sovereignty": each nation should not build its own capabilities but contribute its "unique attributes" to a "coalition of capabilities". One partner supplies critical minerals, another manufacturing, another engineering talent; America supplies the models, the platforms and the cloud, and keeps what Helberg candidly calls "the fat margins, the soaring valuations, the commanding heights of the global economy". This is not a partnership of equals. It is the classic imperial division of labour, updated for the age of artificial intelligence: raw materials and cheap skilled labour flow from the periphery, monopoly rents flow to the centre. At the USISPF summit, Helberg praised India for its "incredible contributions at the application layer". The message could not be clearer. India may build apps on top of American models; it must not presume to build the models themselves.

The numbers show how lopsided this "interdependence" already is. The United States hosts roughly 75% of global AI supercomputer performance and China 15%, leaving the entire rest of the world, Europe included, with barely 10%. US firms plan at least $650 billion in AI-related capital expenditure in 2026 alone, against the European Union's much-trumpeted €47 billion multi-year programme. Any "coalition" built on this asymmetry is not a division of labour among partners. It is a tribute system.

The world has just been shown what happens when the imperial centre decides to pull the plug. On Friday, June 13, at 5:21 pm Eastern Time, a letter from Commerce Secretary Howard Lutnick, threatening criminal penalties under export control law, forced Anthropic to disable its most advanced models, Fable 5 and Mythos 5, for users worldwide. There was no prior warning, no appeals process, no statutory framework, no technical substantiation. Anthropic itself publicly disagreed with the government's reasoning, noting that the cited jailbreak capability existed in other deployed American models, but complied within hours. Access was restored on June 30, as arbitrarily as it was revoked. For seventeen days, governments from Europe to the Gulf to India, which had embedded these models in public services and growth strategies, learned that their access to the "frontier ecosystem" rests on nothing more solid than Washington's discretion. The kill switch is not a metaphor. It has now been used.

China's experience with semiconductors demonstrates that this is not an aberration but the settled policy of US imperialism towards any competitor. Since 2018, successive rounds of American export controls have sought to deny China advanced chips, chipmaking equipment and even the software to design them, with allies like the Netherlands and Japan strong-armed into compliance. The US will not tolerate competition in high technology, whatever the rhetoric about free markets. Yet the sanctions produced the opposite of their intended result: Huawei returned with domestically fabricated advanced chips, Nvidia's AI chip sales in China have stalled as local champions take the lead, and even Washington think tanks now concede the chip war is failing. Technology denial is the imperial weapon; self-reliance is the only proven defence.

India should study its own treatment in this light. India was not even invited to the founding summit of Pax Silica in December 2025; it formally joined the alliance only on February 20, 2026, its "ascension" graciously marked by Helberg's visit to the India AI Impact Summit. At the USISPF gathering, American senators celebrated India as a "highly trusted counterweight" to China, while India's ambassador offered assurances that Atmanirbharta is an ecosystem with "positive externalities". The talk of self-reliance masks a deepening dependence: on American models, American clouds and American goodwill, all revocable by US diktat. A country of 1.4 billion people, with the world's largest pool of engineering talent, is being assigned the role of application-writer and back-office for someone else's intelligence infrastructure.

There is an alternative vision, and it was on display at the UN Open Source Week in New York in the last week of June, where digital sovereignty emerged as the defining theme. Tanzania's minister Angellah Kairuki asked the question Helberg never answers: "Who actually truly owns the ecosystems that serve our people?" For too many nations, she said, the answer had been "a license that we did not write, a platform that we could not inspect, a dependency that we could not break." Tanzania now runs more than 90% of its government systems on open-source software. Cloudera's chief technology officer posed the test every institution must apply to its AI systems: "Can we continue operating if a provider changes its commercial or political position?" After June 13, no honest official can answer yes for the American stack. The consensus of the week was blunt: open source is now core critical infrastructure, and proprietary American companies cannot be trusted with it.

China is championing precisely this open model in AI. DeepSeek's R1, released in January 2025 at a fraction of the compute cost of comparable American models, was released with open weights for anyone to download, inspect and run on their own infrastructure. Leading Chinese open models are now estimated to be only three to six months behind their American counterparts, and the gap is shrinking. The latest release, Z.ai's GLM-5.2, is the most capable open-weight model yet, landing within a few points of Anthropic's Claude Opus 4.8 on long-horizon coding benchmarks at roughly one-sixth the per-token cost of OpenAI's GPT-5.5. Where the American stack offers rented intelligence on revocable terms, the open-source path offers the Global South what it actually needs: models that can be audited, adapted to local languages and needs, and run on sovereign infrastructure that no foreign letter can switch off.

AI is not just another technology. It is fast becoming a core productive force, the operating system of the coming economy, entering production, administration, education and defence. Whoever controls it controls the conditions of development itself. The colonial powers once monopolised industrial machinery and railways; the neocolonial order monopolised finance and intellectual property through instruments like TRIPS. Pax Silica is the attempt to do the same with AI, to bind the world's productive future to American monopoly capital and call the binding "trust".

India and the Global South face a choice that Helberg has helpfully clarified. Either build sovereign and independent AI capability on the foundation of open source, public compute and South-South cooperation, or accept a subordinate berth in an American-run "ecosystem" whose terms were demonstrated on June 13. The digital sovereignty trap is real, but it is not the one Helberg describes. The trap is his offer.