Anthropic v the US Government: AI, free speech and the Fable 5 ban

‍Published by Ross Holmes Lawyers (RHL)

One of the world’s leading artificial-intelligence companies is locked in an extraordinary legal fight with the United States Government — and this month it escalated sharply. Anthropic, the maker of the Claude AI assistant, is suing the Trump administration after being blacklisted by the Pentagon; then, on 13 June 2026, Washington ordered it to block every foreign national on Earth, New Zealanders included, from its most powerful models. This article explains the dispute, the ban, and the bigger question beneath it: how far can a government go in punishing a private company for what its technology will and won’t do? It matters here too, because much of the technology New Zealand businesses rely on rests on foundations a foreign government can move without notice.

Key points

•     Anthropic, maker of the Claude AI, is suing the Trump administration after the Pentagon labelled it a “supply chain risk”.

•     A US federal judge granted Anthropic an injunction, pointing to possible “First Amendment retaliation”.

•     On 13 June 2026, the US barred all foreign nationals — including New Zealanders — from Anthropic’s newest models, Fable 5 and Mythos 5.

•     The fight pits free speech and due process against national-security and export-control powers.

•     For anyone who relies on US technology, it is a reminder that access can change with a single directive.

What happened

‍In March 2026, Anthropic took the unusual step of suing the United States Government. The trigger was a Pentagon decision to designate the company a national-security “supply chain risk” — a label more commonly aimed at firms from foreign adversary states, such as the Chinese technology giant Huawei, and reportedly the first time it had been applied to a US company in this way. The designation did not just stop the military from using Anthropic’s technology; it required defence contractors to certify that they do not use its models at all.

The dispute had been building for months. According to Anthropic’s court filings and contemporaneous reporting, the falling-out followed the company’s refusal to strip long-standing usage restrictions from its government contracts — limits it places on uses such as mass surveillance of citizens and fully autonomous lethal weapons. The US Defence Secretary wanted those restrictions removed; Anthropic declined. The administration then applied the supply-chain-risk label, and the President directed federal agencies to stop using the company’s Claude chatbot.

Anthropic filed two cases — one in the federal court in California, another in the federal appeals court in Washington, DC — arguing the action was unlawful, exceeded the government’s statutory powers, and punished the company for exercising free speech. The Government has defended the designation as lawful and justified, and a White House spokeswoman publicly dismissed Anthropic as a “radical left” company trying to dictate military policy. In late March 2026 a federal judge in San Francisco granted Anthropic a preliminary injunction, questioning the government’s apparent effort to cripple the company and pointing to the possibility of “First Amendment retaliation”. The Washington appeal was argued in May 2026 and, at the time of writing, awaits a decision.

The latest twist: a worldwide ban on Fable 5

‍Then it escalated again. On 9 June 2026, Anthropic released a powerful new model, Claude Fable 5 — a publicly available version of its most capable “Mythos-class” technology, notable (among other things) for its ability to find flaws in software. Four days later, on the evening of Friday 13 June 2026, the US Commerce Department issued an export-control directive suspending access to Fable 5 and the non-public Mythos 5 for any foreign national — whether inside or outside the United States, and including Anthropic’s own non-citizen staff.

Because the ban was so widespread, Anthropic said it had no practical choice but to switch the two models off for everyone, not just foreign users, to ensure compliance. Access to the company’s other models, including Claude Opus 4.8, was unaffected. The Government cited national security, and Anthropic understood that officials believed they had found a way to “jailbreak” Fable 5 — to circumvent its safety limits. Anthropic said it had been given only limited, verbal information, disagreed that a narrow vulnerability justified pulling a product used by hundreds of millions of people, and said it was working to restore access.

The one thing most people get wrong

‍This is not a content-moderation row, nor an ordinary commercial disagreement. It is a constitutional and export-control fight — and that is what makes it so significant. The “supply chain risk” designation at the heart of it is a national-security tool generally reserved for foreign adversaries, now turned on an American company; and the Fable 5 order uses export-control powers — the same legal machinery that governs the sale of sensitive goods overseas — to switch off a commercial software product around the world.

‍That second point is the one that should give every business pause. A single directive from one government, issued early on a Friday evening, was enough to disable a widely used tool for lawful users everywhere — including in New Zealand — with no notice. Whatever the rights and wrongs of this particular case, it is a vivid illustration of how much modern technology sits on foundations controlled by a single jurisdiction.

National security, or payback?

‍Here is the question many are now asking: Is this genuine national-security caution, or retaliation against a company that refused to fall into line? There is material on both sides, and it is worth weighing honestly.

The retaliation reading has real support. A federal judge has already pointed to possible “First Amendment retaliation” in the related blacklisting case; Anthropic’s own pleadings frame the whole sequence as punishment for its stated principles; the supply-chain-risk label arrived after it refused the Defence Secretary’s demands; a competitor reportedly signed its own Pentagon deal within hours of Anthropic being punished; and senior officials have used pointedly political language about the company. On this view, the Fable 5 order looks like the latest move against an organisation that would not simply say “yes, sir”.

The national-security reading is not empty either. Fable 5 is, by Anthropic’s own account, exceptionally capable — the company said at launch that releasing something this powerful carried risks, and built in safeguards precisely because of its abilities, including in cybersecurity. Officials say they have identified a method to bypass those safeguards. And some commentators have made a pointed observation: a company that keeps describing its own product as near-weapons-grade should not be shocked when a government chooses to regulate it as such.

‍Both things can be partly true at once. What is clear is that the courts — not press releases from either side — will ultimately decide whether the Government acted within its powers. For now, reasonable people are drawing different conclusions, and it is fair to keep an open mind.‍ ‍

What it means for New Zealanders

‍You do not need a view on US politics to take a practical lesson from this. A great deal of the software New Zealand businesses and professionals use every day — including AI tools — runs on infrastructure and models controlled from overseas and subject to foreign law. When access can be withdrawn by directive, a few questions are worth asking:‍ ‍

•     Could your business keep operating if a key overseas tool were switched off at short notice?

•     Do your contracts with suppliers address continuity, notice, data access, and what happens if the service becomes unavailable?

•     Are critical processes over-reliant on a single foreign platform, with no fallback?

None of this is a reason to avoid good technology. It is a reason to use it with your eyes open — and to ensure the legal and commercial arrangements surrounding it are sound. ‍

Where to get information and support

‍•     Anthropic’s own newsroom and statement on the directive: anthropic.com.‍ ‍

•     Independent coverage from established outlets such as Reuters, the BBC, CNN and Al Jazeera.

•     The court proceedings themselves, filed in the US federal courts in California and Washington, DC.

One change to watch

This story is moving quickly. Watch for the Washington appeal court’s decision, any restoration of Fable 5 access, and whether the two sides reach a settlement — Anthropic has said litigation does not rule that out. More broadly, the case is shaping how the US uses export controls over advanced AI, which will affect access far beyond America’s borders.

This article is general information only and is not legal advice. It comments on ongoing overseas litigation that may change. Your situation is unique; please obtain specific advice before acting.