While the Trump administration was declaring war on one of its leading artificial intelligence companies last week, its Commerce Department was also circulating drafts of new rules that would regulate exports of the leading-edge semiconductors essential to the development of AI.
Just as the Trump administration’s decision to extend the Pentagon’s declaration of Anthropic as a “supply chain risk” was being extended to other US government agencies in an existential threat to the company at the forefront of deployment of AI tools, the administration is contemplating giving itself the power, on a case-by-case basis, to decide whether countries are given access to the chips that are vital to build the data centres that train and run AI models.
Moreover, while countries would be obliged to ask the White House for permission to access the most powerful chips made by companies like Nvidia and AMD – giving the administration a veto over whether countries can build their own AI infrastructure – they could also be obliged to make “matching” investments in American AI.
In effect, if countries want to participate in a technology that promises to transform economies, they would need America’s permission and could be forced to pay for the privilege.
President Donald Trump, whose use of tariffs to extort investment pledges has been undermined by the US Supreme Court’s declaration that his “reciprocal” tariffs are illegal, may have found another form of leverage to compel other countries to invest in the US.
The administration has been at pains to say that the proposed rules are only a draft, while also saying that it is committed to promoting secure exports of the US tech stack and describing the attempts by the Biden administration to regulate semiconductor exports as “burdensome, overreaching and disastrous”.
The Biden administration banned the export of Nvidia and AMD’s most advanced chips to China, capped the volume of chip sales to most countries and proposed a two-tier system, where exports to 18 “tier one” close allies (including Australia) would be allowed freely, but sales to “tier two” countries regulated and sales to a number of other countries – including China and Russia – would be prohibited.
Trump rescinded that policy before it could take effect (as he did to the Biden administration’s tentative efforts to develop rules governing the safety and transparency of AI development) and his administration’s approach to AI exports has been more transactional, and ad hoc.
Last year, for instance, he decided Nvidia and AMD would be able to export advanced AI chips (albeit, not their most advanced) to China, provided China paid the US government 25 per cent of the revenue generated.
As it transpired, few sales have been made because China, keen to promote its domestic chipmakers and AI sector, directed its bigger tech companies not to buy the US chips.
Trump has also approved sales of hundreds of thousands of the most advanced US chips to Saudi Arabia and the United Arab Emirates – countries that were in Biden’s “tier two” band – in exchange for promises of investment in the US. The Saudis pledged $US1 trillion of investment.
Those deals, while highlighting the transactional nature of the administration, caused controversy once it emerged that a member of the UAE royal family and the chair of its $US1.5 trillion sovereign wealth fund, Sheikh Tahnoon bin Zayed Al Nahyan, had spent $US500 million acquiring a 49 per cent stake in a Trump family cryptocurrency firm in the days leading up to Trump’s inauguration last year.
An entity led by Sheikh Tahnoon also used $US2 billion of a digital currency issued by World Liberty Financial, which is owned by the Trumps and the family of Trump’s Middle East envoy, Steve Witkoff, to acquire a stake in the cryptocurrency exchange, Binance, generating significant revenue for World Liberty Finance.
While Trump says he has no role in running the family businesses, having turned over their management to his children, the switch from Biden’s codified and transparent rules for regulating AI exports to one that could be conducted at the whim of the president underscores the potential for conflicts of interest, even corruption, and uncertainty.
Similar issues are at play in the administration’s fight with Anthropic, whose Claude tools are deeply embedded in the US military and regarded as so critical to the conflict with Iran that they will be phased out over six months.
The administration took violent exception to Anthropic’s “red lines” – it wanted the Pentagon’s assurance that its AI wouldn’t be used for mass surveillance of US citizens (which would be unconstitutional) or for autonomous weapons (“killer robots”) – and not only barred it from working for the Pentagon in future but declared it a supply chain risk, which will deter other companies with military contracts from doing business with it. That declaration was subsequently extended to other government agencies.
This is a designation never before used against a US company – it has been reserved for US foes – and represents a chilling message to other US companies that they either fall into line with the administration or risk existential threats to their businesses.
“Anthropic is in trouble because I fired them like dogs, because they shouldn’t have done that [tried to impose limitations on the use of its products],” Trump said.
Trump has called Anthropic, which has prided itself on prioritising safe use of AI and advocated its regulation, a “radical left, woke company” full of “left-wing nutjobs”.
AI companies keen to retain or win government contracts may not only have to agree to their being used however the Pentagon chooses to, but may also have to be mindful of their diversity, equity and inclusion policies.
The AI sector is incestuous, with layers of financial and technical interconnections across a sector that has an insatiable appetite for capital and where access to capital is both America’s biggest advantage over the rest of the world and the companies’ biggest vulnerability.
Declaring a key player a supply chain risk not only jeopardises the financials stability of a $US380 billion company but, if it is maintained (Anthropic has said it will challenge it in the courts), sends ripples of uncertainty and potential financial stress throughout the sector.
In effect, if countries want to participate in a technology that promises to transform economies, they would need America’s permission and could be forced to pay for the privilege.
The assault on Anthropic has been conducted primarily via social media posts. Which AI company could feel comfortable when a post could destroy their business or, as is the case with Anthropic, force them to remove products deeply embedded in their own systems?
The Trump administration’s desire to micromanage AI, the policies of the companies developing it and the exports from the chipmakers essential to its development, injects unpredictability and risk into the sector.
The discretionary and unpredictable approach it is taking to regulating AI, where which companies can operate and which customers they can sell their services to is seemingly at the whim of an erratic presidency, is at odds with the speed and seamless execution at which AI has to be developed and deployed by the US if it is to justify the trillions of dollar being invested.
Any disruption to that execution and its timelines, or the capital that is funding it, would be traumatic for the sector, US and global financial markets and the US economy.
The Market Recap newsletter is a wrap of the day’s trading. Get it each weekday afternoon.
Stephen Bartholomeusz is one of Australia’s most respected business journalists. He was most recently co-founder and associate editor of the Business Spectator website and an associate editor and senior columnist at The Australian.Connect via email.