Best of Both Sides: US gatekeeping AI will only lead to diffusion

For the world, the move could force AI companies elsewhere to look for models from other countries or build their own. Countries would now want to cooperate more to overcome this American hegemony in frontier AI.

5 min readJul 3, 2026 06:30 AM IST

First published on: Jul 3, 2026 at 06:30 AM IST

America’s governance of AI is proving whimsical, much like the Donald Trump administration’s approach to most other aspects of geopolitics. But its more recent efforts to wall off access to advanced American AI models could end up being counterproductive. Policies designed to further America’s lead could encourage foundational AI model developers worldwide to build on Beijing’s largely open-source model push and unintentionally catalyse China’s rise in AI, while also forcing more cooperation across other jurisdictions such as the EU or Japan. While Trump’s increasingly transactional approach has now been legitimised as state policy, its willingness to treat AI as a new source of extraordinary leverage comes with pitfalls in the long term.

So, how fickle is this new US regulatory stance on AI? A departure from its traditional light-touch regulatory strategy, the new American policy seems to be adjusting and improvising, quite like a saxophone player in a jazz quartet, to the arrival of increasingly powerful models with national-security implications. On June 12, three days after Claude Fable 5 was launched, the Trump administration directed Anthropic to exclude foreigners from accessing its latest and most capable frontier AI model. On Wednesday, after an over two-week ban, the US government allowed Anthropic to restore customer access.

Despite this, the signals these actions send are likely to endure: That America is not loath to exploit its role as the gatekeeper to frontier AI while capriciously deciding who gets to use the world’s most crucial technology. That the US is willing to ride roughshod over private AI companies, as it did earlier in March when it designated Anthropic a supply-chain risk, and it will not hesitate to treat users pretty much the same way. All this despite the models having been developed entirely by private companies with commercial operations across multiple jurisdictions.

While Trump has consistently upended many established tenets of US foreign policy and appears to be unflinchingly proud of brandishing an unreliable image of America in everything from defence to trade, this administration is by no means an exception. After World War II, the US stopped assisting the UK’s nuclear weapons programme, only to relent after the latter developed expertise of its own. It has consistently chosen not to share its best military equipment, like the F-22 fighter jets, even with close allies. Most of this administration’s restrictive policy actions do not just affect America’s adversaries, but also its closest allies.

Reversing the AI ban likely happened not out of altruism, but because of the realisation in Washington DC that the legal basis of the order was shaky and was unlikely to stand up to challenges in courts.

From a strategic standpoint, though, these restrictions seem even more tenuous. Take America’s more recent restrictions focussed on cutting off China. Following export bans that blocked China from accessing advanced semiconductor technologies and space equipment, Beijing has aggressively pivoted to domestic development. In the process, Chinese firms successfully designed and launched their own high-performance chips, specifically testing processors in the 28-nanometre (nm) to 16 nm range, all the way to Huawei’s project aimed at producing 1.4-nm chips by 2031. US space restrictions may have restrained NASA and federally funded contractors such as SpaceX from engaging in collaborations with China, but Beijing has managed to build a top-notch space sector that is achieving historic milestones, including retrieving samples from the far side of the Moon and fabricating its own low Earth orbit space station.

America’s AI move is anti-innovation, too. Innovation works best through diffusion — like GPS becoming dominant because it was widely available, despite it having started as an American military project. A ban on software is even more untenable than Trump’s earlier interventions that blocked access to hardware, such as Nvidia chips.

For the world, the move could force AI companies elsewhere to look for models from other countries or build their own. Countries would now want to cooperate more to overcome this American hegemony in frontier AI. China is already pushing ahead on cost-optimised open-source models and accelerating adoption across its economy; the EU is now fostering cooperation to build sovereign compute while Japan has announced substantial capital investments.

For India, too, it’s becoming clear that without a sovereign AI stack, the reliance on foreign tech could be detrimental, despite all the hype around the country’s diffusion strategy: Hybrid public-private AI diffusion aimed at deploying AI across healthcare, agriculture, and public services. It makes sense for Delhi to double down on committing to its own foundational models, however imperfect, and then leverage its scale to engage multiple ecosystems simultaneously. That should be the singular takeaway from the US actions on Claude.

The writer is National Business, and Explained, Editor, The Indian Express. [email protected]