Two senior Trump administration officials told Reuters Thursday that China’s biggest chipmaker, SMIC, has been quietly shipping semiconductor manufacturing equipment to Iran’s military-industrial complex for nearly a year — and shows no sign of stopping.

The transfers, which the officials described as including technical training on SMIC’s processes, could help Tehran build everything from precision-guided munitions to hardened communications systems at a time when Iranian forces are still exchanging missiles and drones with U.S. and Israeli targets.

The disclosures, made on the condition of anonymity because they involve still-classified intelligence, land like a live grenade in an already combustible mix of Middle East war and U.S.-China rivalry.

Beijing has long insisted its business with Iran is strictly commercial. SMIC itself has repeatedly denied any ties to the Chinese military, even after Washington blacklisted it in 2020 and tightened the screws again in 2024 following its surprise production of a 7-nanometer chip for Huawei’s Mate 60 Pro smartphone.

The timing could hardly be worse for diplomacy. President Donald Trump is scheduled to visit Beijing in May, a trip both sides had hoped would produce at least a framework agreement on stabilizing the bilateral relationship. Negotiators had been eyeing progress on everything from tariff relief to clearer rules on technology exports. Now, the SMIC allegations threaten to poison the atmosphere before Air Force One even touches down.

One U.S. official told Reuters the equipment flow “almost certainly included technical training” on advanced node techniques. While it was not clear whether any of the tools contained U.S.-origin components, which would be a straight-up sanctions violation, the mere transfer of know-how from a company Washington has spent years trying to hobble is being viewed inside the administration as a direct challenge.

SMIC has made incremental but stubborn progress despite export bans on the most sophisticated lithography machines from ASML and U.S. suppliers like Applied Materials, Lam Research, and KLA. It has relied on older deep-ultraviolet tools, multi-patterning tricks, and domestic alternatives to push into 7nm-class production.

Handing even that level of capability to Iran could let Tehran shorten the time it takes to field more sophisticated electronics for its missile, drone, and radar programs — exactly the kind of incremental military edge U.S. and Israeli planners worry about.

The revelations also highlight the limits of America’s long-running campaign to keep China from achieving semiconductor self-sufficiency. Despite layered sanctions, SMIC has kept its most advanced fab in Shanghai humming by sourcing what it can from non-U.S. channels and investing billions in its own R&D. Critics inside the U.S. government have long warned that third-country diversions and dual-use loopholes would eventually undermine the controls; the Iran shipments appear to be Exhibit A.

China, for its part, has walked a careful line since the U.S.-Israeli strikes began on Feb. 28. Foreign Minister Wang Yi repeated this week that all parties should seize any opening for peace talks. Beijing has avoided explicit alignment with either side while quietly expanding its economic footprint in Tehran — including, Reuters reported last month, near-final talks on Chinese anti-ship cruise missiles for Iran.

The SMIC story adds another layer of friction just as markets are still digesting the war’s impact on oil prices and global inflation. Any escalation in U.S. sanctions, whether fresh designations on SMIC executives, secondary penalties on Chinese banks that finance the trade, or new restrictions on allied exports, could ripple through already jittery supply chains. Chip stocks on both sides of the Pacific would feel the chill.

Inside the White House, the disclosures are being read as more than a sanctions enforcement headache, the report indicates. They are seen as a test of whether Beijing is willing to rein in its state-linked champions when those companies cross into active conflict zones.

One official reportedly said that this is not just about chips; it is about whether China will continue to treat sanctioned entities as national champions even when they are feeding the war machine of America’s adversary.

For now, the administration is weighing its next move. Options range from public condemnation to targeted enforcement actions that could complicate the very negotiations Trump hopes to clinch in Beijing.

A U.S. federal judge has intervened in a high-stakes confrontation between the Pentagon and Anthropic, temporarily blocking a move that could have shut the artificial intelligence firm out of lucrative government work and set a far-reaching precedent for how Washington deals with dissenting technology providers.

In a ruling, U.S. District Judge Rita Lin found that the government’s decision to designate Anthropic as a national security supply-chain risk appeared to be driven less by operational concerns and more by retaliation for the company’s public stance on AI safety.

“The record supports an inference that Anthropic is being punished for criticizing the government’s contracting position in the press,” Lin wrote. “Punishing Anthropic for bringing public scrutiny to the government’s contracting position is classic illegal First Amendment retaliation.”

The order, which takes effect after a seven-day pause to allow for an appeal, halts what had been an extraordinary step by the U.S. military establishment. The designation, rarely used and historically aimed at foreign-linked threats, effectively barred Anthropic from participating in certain Pentagon contracts, cutting it off from a fast-growing stream of defense-related AI spending.

The decision by Defense Secretary Pete Hegseth marked the first time a domestic technology company had been publicly labelled a supply-chain risk under the statute, a move that immediately reverberated across Silicon Valley and Washington alike.

How They Landed in Court

At the center of the dispute is a fundamental disagreement over how artificial intelligence should be deployed in military operations.

Anthropic drew a line early, refusing to allow its Claude models to be used for autonomous weapons systems or domestic surveillance. The company has argued that current-generation AI lacks the reliability and alignment safeguards required for lethal or intrusive applications, and that deploying such systems without clear constraints risks both operational failure and civil liberties violations.

The Pentagon sees the issue differently. As AI becomes central to intelligence gathering, targeting, logistics, and cyber operations, defense officials have grown increasingly wary of private companies imposing limitations on how their technology can be used in national security contexts.

In court filings, the Justice Department argued that Anthropic’s refusal to accept certain contractual terms created uncertainty over whether its systems could be relied upon in critical operations. Officials warned that such restrictions could “risk disabling military systems during operations,” framing the issue as one of operational readiness rather than corporate dissent.

Anthropic’s lawsuit presents a sharply different narrative. Filed in California federal court on March 9, it accuses the government of acting unlawfully and without a factual basis, arguing that the designation was inconsistent with the military’s own prior assessments, which had reportedly praised Claude’s capabilities.

The company also claims it was denied due process, saying it was not given an opportunity to contest the designation before it was imposed—an alleged violation of its Fifth Amendment rights.

Beyond the legal arguments, the case is rapidly becoming a test of how far the U.S. government can go in compelling alignment from private-sector AI developers.

For decades, defense contracting has operated on a relatively clear premise: companies that meet technical and security requirements can bid for government work, even if they maintain independent views on policy. The Anthropic case challenges that boundary, raising the question of whether expressing opposition to certain uses of technology can itself become grounds for exclusion.

Anthropic executives have warned that exclusion from defense contracts could cost the company billions of dollars in lost opportunities at a time when government demand for AI capabilities is accelerating. The reputational impact could be just as significant, potentially signaling to other agencies and partners that the firm is politically or operationally contentious.

The Pentagon’s move, and the court’s response, arrive amid a period of intensifying competition among AI developers to secure government contracts, particularly as Washington ramps up spending to maintain technological parity with geopolitical rivals.

For some companies, that has meant leaning into defense partnerships. For others, including Anthropic, it has meant attempting to draw ethical boundaries around how their systems are deployed.

OpenAI quickly stepped in to take the contract after the Pentagon booted Anthropic out. The decision by OpenAI triggered a backlash, resulting in massive uninstalls of its app.

However, Judge Lin’s ruling does not settle the dispute, but it does establish an early check on executive authority. By framing the designation as potential retaliation rather than a clear-cut security measure, the court has signaled that national security claims will not automatically override constitutional protections.

Anthropic, for its part, has struck a careful balance in its public response. “While this case was necessary to protect Anthropic, our customers, and our partners, our focus remains on working productively with the government to ensure all Americans benefit from safe, reliable AI,” spokesperson Danielle Cohen said.

That stance indicates the company is seeking to preserve access to government business while resisting pressure to relax its safeguards—a position that may become increasingly difficult to maintain as defense agencies push for fewer constraints and greater control.

The legal battle is far from over. Anthropic has a second case pending in Washington, D.C., challenging a separate designation that could extend its exclusion beyond the Pentagon to civilian federal agencies. An appeal in the current case could also move the dispute into higher courts, potentially setting a precedent with implications across the technology sector.

What is already clear is that the confrontation marks a turning point.

As artificial intelligence becomes embedded in national security strategy, the relationship between governments and the companies building these systems is shifting—from partnership to negotiation, and in some cases, open conflict. The Anthropic case has brought that tension into the open.