America's shutdown of Anthropic's AI models is not a surprise. And India has seen it coming for 30 years

On June 12, 2026, the United States government pulled what the technology industry has long feared but rarely named aloud: a kill switch. US Commerce Secretary Howard Lutnick signed an export control order that immediately blocked all non-American nationals from accessing two of Anthropic's most advanced artificial intelligence models. Claude Fable 5, which had launched just three days earlier to global acclaim, and Claude Mythos 5, a restricted model whose cybersecurity capabilities had already rattled financial markets and attracted close government scrutiny. Anthropic complied within hours, disabling both models for users outside the United States. India, along with the rest of the non-American world, simply lost access.

The stated trigger was a report that researchers had found a method to bypass the safety guardrails on Mythos 5's cyber capabilities, guardrails designed to prevent the model from being used to attack banks, critical infrastructure, and government systems. Anthropic characterised the issue as minor, noting that similar bypasses could likely be replicated on competing models. The US government was not persuaded. What had been framed for months as a global AI race was, in a single letter, reframed as a national security asset, one that America would control on its own terms.

Also Read| Indian IT industry's mirror to the future develops a ‘serious crack’ in world of restricted access

This Was Not Unpredictable. This Was Inevitable.
India's relationship with transformative technology has followed a recognisable script for three decades. A new technology wave emerges, typically in the United States. Indian professionals and enterprises adopt it rapidly and with great sophistication. Indian institutions, both corporate and governmental, discuss building indigenous alternatives. The discussion does not materialise into sustained investment. The American or Chinese product comes to dominate. And when the geopolitical weather changes, India discovers it has no leverage.

The pattern is not subtle. When the internet arrived in the early 1990s, there was genuine hope that India might produce a competitive browser or portal. It did not. When search engines transformed information access around 1999, the same hope produced the same outcome. When mobile platforms exploded after 2004, when the mobile internet arrived in 2005, when social media went mainstream, when instant messaging became the backbone of daily communication, when cloud infrastructure became essential to every enterprise, at each inflection point, India watched, adapted, and adopted. It built nothing that the world came to depend on.

By 2018, when artificial intelligence began its transition from research curiosity to commercial reality, the pattern was so well established that breaking it required a deliberate act of will. That act of will did not materialise at scale. Today, in 2026, India's AI market is substantially captured by American large language models, American AI enabled search, and American developer toolchains, with Chinese open source models filling whatever gaps remain. Not a single Indian AI product or platform is in the global conversation.

What the Fable 5 Shutdown Actually Tells Us
The shutdown of Fable 5 and Mythos 5 is instructive not because it is dramatic, but because it is mundane. No missiles were fired. No diplomatic incident was declared. A commerce secretary signed a letter, a technology company complied, and millions of users across the world lost access to tools they had integrated into their professional workflows. The mechanism was entirely routine, the routine application of export control law to a new category of strategic asset.

This is precisely what makes it alarming. Export controls on semiconductors, on encryption technology, on satellite components, these have existed for decades. What is new is that the most economically significant and strategically consequential technology of the current era has now been explicitly placed in the same category. The United States has declared, through action rather than rhetoric, that advanced AI is a national security asset it will not share freely with the world.

The implications extend well beyond Anthropic. Google's Gemini, OpenAI's GPT family, Meta's open source models, and Microsoft's Copilot, all of these rest on American corporate and legal infrastructure. All are, in principle, subject to the same export control architecture that just silenced Fable 5. If the geopolitical temperature rises further, over Taiwan, over trade, over any number of potential flashpoints, that architecture can be applied more broadly and more rapidly than most Indian organisations have modelled.

The Question of Strategic Autonomy
India is not without options. The question is whether the current moment will produce a serious policy response or another round of well-intentioned discussion followed by inaction.

There are things that can be done immediately. Organisations across the public and private sectors can accelerate their adoption of smaller, locally deployable models, both domestic offerings and capable open source models that do not require ongoing connectivity to American servers. These alternatives are not as capable as Fable 5. They are, however, available. And right now, availability matters more than marginal performance.

There are things that will take time but cannot be deferred. Indigenous AI research and development requires not only funding but infrastructure, compute clusters, curated datasets in Indian languages, and sustained institutional commitment measured in years rather than quarters. The GPU supply chain is itself constrained by American export controls, which means that even the hardware needed to train competitive models at scale is not freely obtainable. This is a genuine constraint, not an excuse. The answer is not to abandon indigenous development but to pursue it through alternative architectures: more efficient training methods, smaller specialised models, and collaborative research structures that pool compute across institutions. Companies like Zoho and Sarvam AI are already doing this. They should not remain exceptions.

There are things that require political courage. India's corporate technology sector has, with honourable exceptions, been comfortable as a service provider to American technology platforms rather than as a builder of platforms the world depends on. The incentives have been clear and immediate: building for American clients pays well, quickly, and with limited risk. Building world-class products requires capital, patience, tolerance for failure, and a domestic market willing to back indigenous products even when foreign alternatives are initially superior. None of these conditions are easily legislated into existence, but the policy environment, procurement preferences, R&D tax incentives, sovereign compute infrastructure, can shift the incentive structure.

A Note on What This Is Not
This argument is not anti-American. The United States has every right to treat its most advanced AI systems as national security assets. It would be surprising if it did not. This is not a criticism of Anthropic, which was complying with a legal order from its home government. This is not a call for technological autarky. The notion that India should cut itself off from global technology ecosystems is neither feasible nor desirable.

What this is, plainly stated, is a call to stop pretending that dependence on foreign technology infrastructure is a sustainable national strategy. The kill switch has now been demonstrated to exist and to work. Every Indian policymaker, corporate board member, and technology leader should read that demonstration clearly and plan accordingly.

The Thirty-Year Reckoning
In 1993, when the internet arrived, India's technology sector was young and resource-constrained. The failure to build a competitive browser or portal was understandable. In 2026, India is the world's most populous nation, the fifth-largest economy, and home to one of the largest concentrations of technical talent on earth. The continued failure to build anything that the world depends on in the AI era is no longer understandable. It is a choice and a costly one.

The Anthropic shutdown should function as what strategists call a forcing function: an external event that makes the cost of inaction impossible to ignore. India has a satellite navigation system that remains incomplete. It has cloud infrastructure that is substantially foreign-owned. It has AI capabilities that depend almost entirely on American or Chinese models. If any of these dependencies are exercised against India in a moment of geopolitical stress, the options are narrow. The time to build alternatives is not during that moment. It is now.

Genuine globalisation, the kind that offered equal access to global technology as a matter of commercial right rather than geopolitical favour, is over. The countries that understand this earliest will have the most time to build resilient alternatives. India has already lost three decades. The question is whether the moment a commerce secretary's letter went out on June 12, 2026, will finally be the moment India decided to build in earnest, or whether this too will become another entry in the long list of warnings that were noted and forgotten.

The kill switch exists. It has been used. It will be used again. The only question is whether India will be ready.

Pravin Kaushal is director-Mrikal (AI/Data Center) and a young alumni member, Government Liaison Task Force, IIT Kharagpur