The Trump administration has acknowledged that U.S. agencies moved to restrict the use of Anthropic’s artificial intelligence products after the company resisted Pentagon demands over military applications of its Claude chatbot but denied allegations of retaliation against the company.

In a court filing submitted on Monday, the U.S. Department of Justice denied allegations that the administration unlawfully retaliated against Anthropic. However, the filing also confirmed a central element of the company’s case: that federal agencies began cutting off access to Anthropic’s products following disagreements over how the technology could be deployed for military purposes.

The dispute has emerged as one of the most significant legal confrontations between Washington and the rapidly growing AI industry, arriving at a moment when the White House is simultaneously seeking greater oversight of advanced AI systems while encouraging private-sector investment in the technology.

Anthropic, one of the world’s leading AI developers and a rival to OpenAI, filed suit on March 9 against President Donald Trump and Defense Secretary Pete Hegseth, alleging that it was effectively blacklisted after refusing to modify safeguards built into its Claude models.

At the heart of the dispute is Anthropic’s decision to maintain restrictions preventing its AI systems from being used for autonomous weapons or domestic surveillance. According to the lawsuit, those safeguards put the company at odds with Pentagon officials who sought broader military applications for the technology.

The Justice Department argued in Monday’s filing that the government’s actions do not amount to unlawful retaliation and challenged the case on procedural grounds. Government lawyers contended that the restrictions are not subject to judicial review because Anthropic is not contesting a “final agency action.”

That argument could prove pivotal because if accepted by the court, it could allow federal agencies greater discretion in limiting access to government contracts or procurement opportunities without immediate judicial scrutiny.

Yet the filing also appears to support Anthropic’s broader narrative that the restrictions were linked to a policy disagreement rather than technical or security shortcomings.

The Pentagon had imposed a formal supply-chain risk designation on Anthropic, a classification that limited the use of the company’s technology across parts of the federal government. Two sources previously told Reuters that Anthropic’s technology had been used in military-related operations involving Iran, making the company increasingly important to national security agencies.

According to Anthropic’s lawsuit, the designation was imposed after the company declined to remove safeguards designed to prevent its AI systems from being used in autonomous weapons programs and domestic monitoring activities. The confrontation comes as governments worldwide are wrestling with a fundamental question: who ultimately controls the deployment of increasingly powerful AI systems?

The Trump administration has recently moved toward a more active role in overseeing advanced AI. OpenAI has already indicated it will cooperate with a White House initiative that gives federal authorities the ability to evaluate frontier AI models before public release.

That broader push has fueled concerns among some technology companies that government oversight could gradually evolve into pressure over how AI systems are designed, what safeguards they contain, and which applications they support.

Anthropic has framed the issue as a constitutional dispute, arguing that the government’s actions violate its free speech and due process rights. The company is seeking court orders preventing federal agencies from enforcing the designation and blocking any effort to place it on a broader national security blacklist.

The stakes extend well beyond a single company.

Given the potential impact of the outcome, the case has gathered huge interest from the AI industry as it is expected to establish an important precedent on whether AI developers will retain authority over the permissible uses of their models or whether governments can compel changes when national security interests are involved.

It also arrives during a period of growing political debate over AI governance. Some policymakers have argued that frontier AI companies should have greater obligations to support national security initiatives, while others warn that forcing developers to weaken safety protections could create significant risks.

Anthropic has consistently taken a position as one of the industry’s strongest advocates for AI safety. The company has repeatedly warned about the dangers of advanced AI systems being deployed without adequate safeguards and has called for stronger oversight of frontier models. Critics, however, have argued that some of those safety positions could also provide competitive advantages by slowing rivals or shaping regulatory standards.

The legal battle is unfolding as Anthropic prepares for its next phase of growth. The company disclosed on June 1 that it had confidentially filed paperwork for a U.S. initial public offering, potentially setting the stage for one of the largest AI listings in history.

Meanwhile, a second lawsuit remains pending in Washington, D.C., involving another Pentagon supply-chain risk designation that could ultimately affect Anthropic’s eligibility for civilian government contracts.

Together, the two cases are likely to become closely watched tests of the relationship between Washington and the AI industry.

The United States is escalating efforts to curb China’s access to advanced technology, with lawmakers seeking tighter restrictions on semiconductor manufacturing while the Pentagon broadens its list of Chinese companies deemed connected to Beijing’s military apparatus.

The twin developments highlight how national security concerns continue to shape Washington’s technology and trade policies toward China, even as both countries attempt to stabilize broader economic relations following recent summit between President Donald Trump and Chinese President Xi Jinping.

The latest push is buoyed by growing concern in Washington that existing export controls may still leave avenues for Chinese firms to obtain advanced artificial intelligence chips and computing capabilities through overseas subsidiaries or third-party arrangements.

A bipartisan pair of senators, Republican Jim Banks and Democrat Andy Kim, has urged the Trump administration to tighten regulations governing contract chip manufacturers such as Taiwan Semiconductor Manufacturing Company. Their concern is that Chinese companies could potentially circumvent existing restrictions by using overseas affiliates or intermediary entities to commission advanced AI chips.

The lawmakers’ intervention follows recent guidance from the U.S. Commerce Department’s Bureau of Industry and Security (BIS), which clarified that exports of advanced semiconductors to overseas subsidiaries of Chinese companies require licenses. The clarification addressed concerns that emerged after Washington stepped back from enforcing certain global chip access rules introduced under the previous administration.

However, export-control specialists argue that another vulnerability remains. Chinese firms could potentially use front companies or intermediaries to order custom-designed chips from leading foundries, including TSMC, without directly violating existing restrictions.

In their letter to BIS chief Jeffrey Kessler, the senators warned that failing to close such gaps could undermine the effectiveness of broader U.S. technology controls.

“Should this gap remain unaddressed, ?it would substantially undermine every other restriction the United ?States has ?imposed on the (China’s) access to advanced placed computing capability,” the senators wrote. “Export controls that can be circumvented through fabrication orders at the world’s most advanced foundry ?offer ?no meaningful protection to American national security or ?to the competitiveness of United States industry.”

The issue bears heavy weight because TSMC remains the world’s most advanced contract chip manufacturer and plays a critical role in producing cutting-edge semiconductors used in artificial intelligence systems, data centers, military applications, and advanced computing.

Pentagon widens scrutiny of Chinese technology firms

The lawmakers are pushing the concern as the Pentagon significantly expands its list of companies it considers affiliated with China’s military or defense-industrial ecosystem.

Among the most prominent additions are Chinese technology giants Alibaba Group and Baidu, alongside electric vehicle manufacturer BYD. The updated “1260H list” also includes several firms operating in strategically important sectors such as semiconductors, biotechnology, robotics, and advanced manufacturing.

While inclusion on the list does not trigger direct sanctions, it carries important consequences. The U.S. Department of Defense will be prohibited from contracting directly with listed companies beginning later this month. Restrictions will expand further in June 2027, when procurement through third parties will also be barred.

Analysts note that these indirect restrictions can have meaningful commercial implications because defense contractors and suppliers often adjust procurement practices to avoid compliance risks.

“These indirect restrictions could force some U.S. firms that work with the U.S. military to drop designated Chinese firms as suppliers,” said Michael Hirson, head of China Research at 22V Research.

The Pentagon’s move bolsters a broader U.S. view that China’s civilian technology sector and military modernization efforts are increasingly interconnected through Beijing’s “military-civil fusion” strategy.

The expanded list now stretches across a wide spectrum of industries, from internet platforms and electric vehicles to biotechnology and robotics. New additions include memory-chip manufacturers CXMT and YMTC, biotech company WuXi AppTec, lidar producer RoboSense Technology, and robotics manufacturer Unitree Robotics.

The inclusion of Unitree is notable because the company has attracted international attention for its humanoid robotics technology and recently announced collaborations involving AI research initiatives.

Chinese firms push back

Several companies named on the list have strongly rejected the Pentagon’s characterization. Alibaba said there was no basis for its designation and argued that it is neither a military company nor part of any military-civil fusion initiative. The company indicated it would pursue all available legal avenues to challenge the decision.

“There’s no basis to conclude that Alibaba should be placed on the Section 1260H List. Alibaba is not a Chinese military company nor part of any military-civil fusion strategy. We will take all available legal action against attempts to misrepresent our company,” the company said in a statement to CNBC.

Baidu similarly rejected the designation and pledged to seek removal from the list.

Electric vehicle maker NIO, another newly listed company, said the procurement restrictions would not materially affect its business operations but that it would engage with U.S. authorities to contest the designation.

Their responses are not without precedent. Chinese smartphone manufacturer Xiaomi successfully challenged a previous Pentagon designation in a U.S. court and secured its removal from the list in 2021.

China’s Foreign Ministry also criticized the latest actions, accusing Washington of using national security concerns as a pretext to discriminate against Chinese companies and pledging to protect the interests of affected firms.

Hardware used to be brutal.

Producing a hardware product required enormous funding, long lead times, and the sort of risk that scares off most entrepreneurs. The statistics confirm as much– CB Insights reports that 97% of hardware startups are late to market with their product, and 70% never launch at all.

But something has changed.

Print-on-demand manufacturing has changed the game for hardware entrepreneurs. Going from something that took half a year and six figures, to just days for the cost of a nice laptop. It’s also changing who is able to make hardware at all.

What’s Inside This Guide:

• Why Hardware Used To Be A Founder’s Worst Nightmare
• How On-Demand Fabrication Changed The Game
• The Real Cost Savings For Modern Hardware Startups
• 4x Ways To Use On-Demand Fabrication In Your Startup

Why Hardware Used To Be A Founder’s Worst Nightmare

Let’s rewind for a second.

Ten years ago to validate an idea as a hardware founder, you needed deep pockets. Tooling costs alone were $50,000+. MOQ’s meant founders had to order thousands of units before their first customer paid them anything.

If you were a startup using custom steel parts for your chassis/bracket/enclosure/structural part, this was rough. Shops specializing in custom metal parts manufacturing barely existed in their current incarnation — you hunted down local fabricators, got quotes, got turned down for low volume, and repeated.

Here’s the problem with that old model:

• Massive upfront capital
• Lead times measured in months, not days
• High minimum order quantities
• No room to iterate without burning cash
• Tooling locked you into one design forever

If your prototype was incorrect, they threw away the entire project. Which they typically did – that’s prototyping for you.

No wonder so many killer hardware ideas got sketched out on a whiteboard then died. It was too risky. The stakes of being wrong were just too high. Entrepreneurs weren’t competing on innovation, they were competing on bank accounts.

How On-Demand Fabrication Changed The Game

Here’s where things get interesting.

On-demand manufacturing allows entrepreneurs to upload a CAD file and have a completed part shipped back to them, typically within days. No tooling. No MOQs. No factory relationships. No negotiating with offshore suppliers. One quote. Click. Part.

Need ten pieces? Purchase ten. Want to redesign and purchase ten more next week? No problem. Advanced job shops offer laser cutting, bending, welding, and finishing of parts that once required a supplier partnership just to quote.

That change is also reflected in hard data. Market research firm QY Research valued the on-demand manufacturing market at $5.97 billion in 2024 alone, predicting it will reach $16.68 billion by 2031.

That growth is being driven by startups — not just big factories.

The Real Cost Savings For Modern Hardware Startups

Founders who use on-demand fabrication save in three big areas. Let’s break them down.

Time

Old way: 8 to 12 weeks for a tooled part to arrive.

New way: 3 to 7 days for finished custom steel parts.

That’s not an incremental improvement…That’s going from three iterations per year to thirty iterations per year. Velocity is exponential.

Money

No tooling costs. No deposits. No MOQs. Founders only need a few thousand dollars to launch. Not a few hundred thousand. That means who gets to be a hardware founder in the first place changes.

Risk

If ten dollars worth of parts costs about the same per unit as ten thousand dollars worth, entrepreneurs aren’t forced to invest all their capital into one design. They can:

• Test multiple prototypes side by side
• Get real customer feedback before scaling
• Pivot the design without losing their shirt

That last bit is the important part. Hardware founders don’t have to be right on first try.

4x Ways To Use On-Demand Fabrication In Your Startup

Here are just some of the ways you can integrate on-demand fabrication into your hardware startup. These are the four methods that pack the biggest punch.

Rapid Prototyping

The most obvious use case… And still the most powerful.

Founders can place an order Monday, receive it Friday, prototype it over the weekend, and order version 2 Monday. That quick iteration is what distinguishes teams that launch from those that flounder.

CB Insights reported that 42% of startups fail due to no market need for their product. Rapid prototyping allows founders to validate market fit before going all in.

Small-Batch Production

You don’t need a factory to sell your first 50, 100, or even 500 products.

On-demand shops enable small batches to be produced at prices that actually make sense for early sales. This means founders can:

• Sell to early customers before raising a big round
• Validate pricing and real demand
• Build a revenue track record
• Avoid raising money just to fund inventory

This is huge for bootstrappers and angel-backed startups.

Custom Replacement Parts

Here’s something most hardware founders don’t think about until it bites them…

Spare parts. Repairs. Customer service replacements.

Without on-demand fabrication, founders hoard piles of spare inventory, “just in case.” With it, they make custom steel parts as repairs arrive — cutting inventory costs and unlocking capital for growth.

Limited-Edition Or Bespoke Runs

This one is sneaky.

With on-demand manufacturing, it’s simple to produce limited editions, white-label versions, or bespoke variants of your product without tooling up again. Here’s how hardware founders can leverage it:

• Test new markets without commitment
• Offer premium tiers at premium prices
• Run partnership editions with other brands
• Build a sense of exclusivity around their product

Big margins, small risk.

Bringing It All Together

Hardware isn’t just for founders with wealthy uncles or billions of dollars in seed money anymore. On-demand manufacturing has reduced the barrier to entry in a way that is quietly revolutionizing the industry.

To quickly recap:

• Hardware startups used to need huge upfront capital
• On-demand fabrication eliminates tooling, MOQs, and long lead times
• Founders save time, money, and risk on every iteration
• Custom steel parts can be ordered in batches of 1, 10 or 1,000
• Small teams can finally compete with established brands

The economics have changed. Founders who embrace on-demand manufacturing will iterate quicker, ship earlier, and spend less capital along the way. Those who don’t… will continue to lose ground to those who do.

Bottom line. If you are developing a physical product in 2026, on-demand manufacturing is no longer a luxury…it’s table stakes for your business model.