Anthropic is not done fighting the Pentagon’s “supply chain risk” label. On Tuesday, October 13, 2026, the AI company asked the U.S. Court of Appeals for the District of Columbia Circuit to rehear its challenge to the Department of Defense’s designation, less than three weeks after a divided panel sided with the government. The petition, reported first by CNBC, sets up one of the most closely watched legal fights in the AI industry: whether a federal supply-chain law written to target hostile foreign actors can be used against a US-based AI lab that refused to let its models be used for mass surveillance or autonomous weapons.
The case matters well beyond Anthropic’s legal bills. It tests how far the federal government can reach into a private company’s product decisions under national-security authority, and it lands at a moment when Washington, AI labs, and the Pentagon are all racing to define the rules for military use of large language models. For anyone tracking the AI-models cluster this week, the rehearing petition is the story to watch.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Anthropic Is Asking the Court to Do
Anthropic’s rehearing petition does not ask the D.C. Circuit to simply reconsider a technicality. It asks the full appeals court, or at minimum the original three-judge panel, to revisit a 2-1 ruling issued on September 25, 2026, that upheld the Pentagon’s designation of Anthropic as a national-security supply-chain risk. According to CNBC’s reporting on the filing, Anthropic’s lawyers argue the panel majority effectively rewrote the underlying statute, turning a tool meant for “hostile foreign actors” into one that, in Anthropic’s words, “presumptively reaches every government contractor.”
That framing is the crux of the whole dispute. The relevant law, 41 U.S.C. § 4713, exists to let federal agencies screen out suppliers tied to adversarial foreign governments or compromised hardware and software supply chains. Anthropic’s position is that applying that authority to a US-headquartered AI developer, over disagreements about how its own models should be used, stretches the statute far past what Congress intended when it was written. The Pentagon and the D.C. Circuit majority disagree, and that disagreement is now headed back to the same court for a second look.
Inside the September 25 Ruling: A 2-1 Split
The panel decision that triggered this petition was not unanimous, and that split is exactly what Anthropic is leaning on. Two of the three judges sided with the Department of Defense, while one dissented, according to reporting from UPI and Politico on the September 25 ruling. The majority concluded, in the court’s own language, that the Pentagon “had ample support for its conclusion” that Anthropic’s AI products posed a supply-chain risk.
A divided panel is not unusual in appellate litigation, but it does give a losing party a concrete argument for rehearing: if one of three judges on the original panel saw the law differently, the case may be exactly the kind of close call that benefits from either an en banc review by the full bench or a second look by the same panel. Anthropic’s petition leans into that disagreement rather than treating the loss as settled.
The Statute at the Center of the Fight
41 U.S.C. § 4713 is the specific provision the D.C. Circuit majority cited in its ruling. It is part of the federal government’s broader supply-chain-risk toolkit, the same general category of authority that has historically been used against hardware and telecom suppliers with ties to foreign adversaries. Anthropic’s attorneys argue that stretching this authority to cover a domestic AI lab’s product and safety decisions is a meaningfully different use of the law than Congress had in mind, and that the panel majority’s reading effectively removes the “hostile foreign actor” limitation altogether.
The Pentagon’s Supply Chain Risk Designation, Explained
The underlying dispute traces back to March 2026, when the Department of Defense designated Anthropic a national-security supply-chain risk, according to reporting cited by UPI and Politico. The exact date the designation was issued within that month was not established in available reporting, and the Pentagon’s full stated rationale has not been published in detail. What is confirmed is that the designation placed restrictions on how Anthropic’s Claude models could be procured or used by the department, and it set off a legal fight that has now run for more than seven months.
Notably, this was not the first time a court weighed in. The New York Times reported in March 2026 that a judge stayed the Pentagon’s labeling of Anthropic as a “supply chain risk,” meaning the designation faced legal pushback almost immediately after it was issued. That early stay makes the September 25 appeals court ruling, which effectively lifted or upheld the designation over the stay, a significant reversal of fortune for Anthropic in the litigation.
Why Anthropic Says the Designation Was Retaliation
The most pointed argument in Anthropic’s petition is not about statutory interpretation. It is about motive. Anthropic’s attorneys wrote that the court allowed the Pentagon to act “without any pre-designation process and despite straightforward evidence the designation was retaliation for Anthropic’s speech about AI safety,” according to the filing details reported by CNBC. That is a direct accusation that the national-security label was less about actual supply-chain risk and more about punishing Anthropic for its public positions on how AI should and should not be deployed by the military.
Specifically, Anthropic’s lawyers said the company was branded a national-security risk because it declined to permit its models to be used for “mass surveillance of Americans” or “lethal autonomous warfare,” according to the petition as described by reporting tied to the filing. If that characterization holds up under judicial scrutiny, it reframes the entire case: not as a routine supply-chain compliance dispute, but as a test of whether a federal agency can use national-security authority to pressure a contractor into loosening its own safety guardrails.
This argument also connects to a broader pattern industry watchers have flagged around AI labs and government pressure campaigns. Anthropic has separately disclosed friction with state actors attempting to misuse its models, including a case in which the company said Russia attempted to weaponize Claude for killer drones. Whether that history helps or hurts Anthropic’s retaliation argument in court is an open question, but it underscores how central the “what should AI be allowed to do in warfare” debate has become to the company’s regulatory posture.
Case Timeline: From Designation to Rehearing Petition
The dispute has moved through multiple stages in just over seven months. The table below lays out the confirmed sequence of events based on verified reporting.
| Date | Event | Outcome |
|---|---|---|
| March 2026 | Department of Defense designates Anthropic a national-security supply-chain risk | Restrictions placed on Claude procurement/use by DoD |
| March 26, 2026 | Federal judge issues a stay on the Pentagon’s labeling of Anthropic | Designation temporarily blocked, per The New York Times |
| September 25, 2026 | D.C. Circuit panel rules 2-1 in favor of the Pentagon | Designation upheld; one judge dissents |
| October 13, 2026 | Anthropic files rehearing petition with the D.C. Circuit | Case pending; outcome unconfirmed |
Market Impact: What’s at Stake for Anthropic’s Government Business
Available reporting does not establish a dollar figure for business Anthropic has lost or stands to lose because of the designation, and this article will not speculate one. But the strategic stakes are easy to infer from the facts on record. A standing national-security supply-chain risk label makes it far harder for any federal agency, not just the Pentagon, to procure or integrate Claude models, since such designations typically trigger broader federal contracting scrutiny beyond the single agency that issued them.
That matters because federal and defense-adjacent contracts have become an increasingly visible part of how major AI labs talk about growth. Anthropic has publicly pursued enterprise and government deployments of Claude as a pillar of its commercial strategy, alongside its work on cybersecurity defense tooling such as Scion, its internal source-control and code-review system. A durable “risk” label, even one under active appeal, creates a chilling effect: procurement officers tend to avoid vendors with contested national-security designations regardless of where the litigation ultimately lands, simply because the compliance risk of being wrong is asymmetric.
There is also a reputational dimension that extends past government contracts. Anthropic has built its public identity substantially around AI safety positioning, and a federal court record stating the Pentagon “had ample support” for treating the company as a security risk is a headline that competitors, lawmakers skeptical of AI, and even enterprise customers evaluating vendor risk will all have in hand. Anthropic’s decision to keep fighting, rather than settle into the designation, suggests the company sees that reputational exposure as more costly than the legal fees of a rehearing push.
Competitive Landscape: How AI Labs Navigate Defense and Government Work
Anthropic is far from the only major AI lab wrestling with how to position itself for federal and defense-sector business while maintaining safety commitments. OpenAI, Google, and Microsoft have all pursued government AI contracts in 2026, and all three have faced some version of the same tension Anthropic is litigating: federal buyers want capable models, but they also want assurances those models won’t be weaponized or misused in ways that create legal and political blowback for the agencies deploying them.
What sets Anthropic’s situation apart is that its dispute escalated into formal supply-chain-risk litigation rather than staying a private contracting disagreement. That is a meaningfully different outcome than, for instance, the model-access restrictions the White House has placed on firms in other contexts, such as the limits reported on AI model access for UK testers, which involved export-style controls rather than a formal risk designation. Anthropic’s case is now a live test of whether declining a government use case, rather than failing a technical security review, can itself trigger the most severe classification the federal supply-chain framework offers.
Rival labs are watching closely because the legal theory the Pentagon is using, if it survives rehearing and any further appeal, would not be limited to Anthropic. Any AI company that places restrictions on surveillance or autonomous-weapons use cases, policies that are now standard across most frontier labs’ public usage policies, could in theory face the same kind of designation. That is effectively the argument Anthropic is making to the court: this is not a one-company problem, it is a precedent-setting one.
How Rivals’ Usage Policies Compare
Most frontier labs, including OpenAI and Google, publish usage policies that restrict or prohibit mass-surveillance and autonomous-weapons applications in some form, similar in spirit to the restrictions Anthropic cites in its own defense. That similarity is exactly why Anthropic’s rehearing petition carries weight beyond its own four walls: if the Pentagon’s theory survives, any lab with comparable restrictions could face the same exposure the next time a federal buyer wants a use case the lab’s own policy rules out.
Historical Context: Supply Chain Risk Designations Before AI
Supply-chain-risk authority under federal law has historically been aimed at hardware and telecommunications suppliers with ties to foreign governments the US considers adversarial, rather than at domestic software companies disputing their own product policies. Reports on the Anthropic case, including coverage referenced by Congress’s own research arm, frame this dispute as a novel application of that authority to a US-based AI developer, which is part of why the legal theory is drawing so much attention from policy watchers.
Congress has taken enough notice of the Pentagon-Anthropic dispute to produce formal analysis through its research service, a sign that lawmakers see this as more than an isolated contracting disagreement. The Congressional Research Service published an “In Focus” brief specifically on the Pentagon-Anthropic dispute over autonomous weapon systems, flagging it as a potential issue for Congress to address directly, rather than leaving the question entirely to the courts. That alone signals this case could shape how Congress eventually writes AI-specific procurement and national-security rules, instead of leaving agencies to stretch decades-old supply-chain statutes to cover frontier AI products.
Industry and Policy Reaction
The case lands amid a broader wave of scrutiny over how AI companies interact with federal power. Just weeks before Anthropic’s rehearing petition, the Trump administration moved several AI policy officials associated with the Commission on AI Safety and Security Infrastructure into the White House, a reshuffling covered in reporting on the move of CAISSI officials to the White House. That kind of structural change in who sets federal AI policy adds another layer of uncertainty to how the Pentagon dispute could evolve, since the officials shaping future guidance may not be the same ones who backed the original designation.
Separately, concerns about insider risk and security vetting inside major AI labs have been a recurring theme in 2026 coverage, including unrelated but thematically adjacent reporting on insider risk exposure at AI labs. While that story involves a different company and a different set of facts, it reflects the same underlying anxiety driving the Pentagon’s posture toward Anthropic: a sense among federal security officials that AI labs, as a category, need tighter national-security oversight than traditional software vendors.
Why Congress Is Paying Attention
A Congressional Research Service brief treating the dispute as a potential legislative matter, rather than a purely judicial one, suggests some lawmakers worry the courts alone cannot settle how national-security supply-chain law should apply to AI. That kind of attention typically precedes hearings, letters to agency heads, or draft legislation, any of which could reshape the legal landscape around this case well before the D.C. Circuit or a higher court issues a final word.
What’s Confirmed and What Remains Unknown
Given how fast this case is moving, it is worth being explicit about which details are nailed down and which are still open. The table below separates verified reporting from the questions that remain unanswered as of October 14, 2026.
| Topic | Status | Detail |
|---|---|---|
| Rehearing petition filed | Confirmed | Filed October 13, 2026, with the D.C. Circuit |
| Panel ruling margin | Confirmed | 2-1 decision issued September 25, 2026 |
| Governing statute | Confirmed | 41 U.S.C. § 4713 cited by the majority |
| Exact designation date | Unconfirmed | Only “March 2026” established in reporting |
| Number/versions of Claude models covered | Unconfirmed | Not specified in available reporting |
| Dollar value of lost business | Unconfirmed | No figure established in available reporting |
| Rehearing outcome | Unconfirmed | Court has not ruled on the petition |
What Happens Next: The Odds of a Rehearing
Rehearing petitions, whether sought from the same three-judge panel or the full D.C. Circuit sitting en banc, are granted far less often than they are filed. Appeals courts generally reserve full-court rehearings for cases involving a genuine split in legal reasoning across the circuit or an issue of exceptional importance, and a 2-1 panel decision, on its own, is not automatically enough to clear that bar. What works in Anthropic’s favor is the dissent itself: a disagreement among the original judges gives the full court a documented internal split to point to if it wants to take the case up again.
What works against Anthropic is the broader posture of appellate courts toward reversing panel decisions on novel statutory questions, especially ones framed as deferring to executive branch national-security judgment. Courts have historically been reluctant to second-guess an agency’s stated security rationale, even when the underlying facts are contested, which is part of why the majority’s “ample support” language is likely to be hard to dislodge without new evidence or a clear legal error in the panel’s reasoning.
Predictions: Where the Anthropic-Pentagon Fight Goes From Here
Based on the confirmed facts and the posture both sides have taken, a few outcomes look more likely than others over the coming months.
- The full D.C. Circuit likely declines en banc rehearing but may ask for supplemental briefing. Given how procedurally unusual full-court rehearings are, expect the court to either deny the petition outright or request narrower briefing on the retaliation claim specifically, rather than reopening the entire case.
- Congress moves faster than the courts on a legislative fix. With the Congressional Research Service already flagging this as a potential congressional issue, expect lawmakers to introduce or advance AI-specific supply-chain or procurement legislation before the litigation fully resolves, which could moot parts of the case.
- Other AI labs stay publicly quiet but privately lobby. Expect OpenAI, Google, and Microsoft to avoid direct public comment on Anthropic’s case while increasing behind-the-scenes engagement with defense procurement policy, since the legal theory could eventually touch their own usage-policy restrictions.
- Anthropic continues pursuing non-DoD federal and enterprise deals regardless of the appeal’s outcome. The company has shown no sign of retreating from government-adjacent work, including cybersecurity defense products, and is likely to keep expanding there while the Pentagon matter plays out in parallel.
- The case becomes a reference point in future AI-defense disputes. Whether Anthropic wins or loses the rehearing, expect the ruling, and the retaliation argument specifically, to be cited in future disputes between AI developers and federal agencies over usage-policy enforcement.
What This Means for Enterprise and Government AI Customers
For procurement teams at agencies and large enterprises that already use Claude, the immediate practical impact is limited mostly to uncertainty rather than any confirmed new restriction outside the Defense Department itself. But uncertainty has its own cost. Federal contracting officers tend to be risk-averse, and a published appeals court ruling describing a vendor as having “ample support” behind a national-security risk label is the kind of language that shows up in due-diligence memos for years, regardless of how the rehearing turns out.
Enterprise buyers outside government procurement face a different calculus. Most are more concerned with whether a vendor’s legal exposure could disrupt product roadmaps or support commitments than with the national-security specifics. On that front, Anthropic’s continued investment in commercial products, from frontier model releases like Claude Opus 5.5 to enterprise safety tooling, suggests the company is treating the Pentagon dispute as a contained legal fight rather than an existential threat to its broader business, even as it fights the designation hard in court.
Frequently Asked Questions
What did Anthropic file on October 13, 2026?
Anthropic filed a petition asking the U.S. Court of Appeals for the District of Columbia Circuit to rehear its challenge to the Pentagon’s designation of the company as a national-security supply-chain risk, according to CNBC’s reporting on the filing.
What was the September 25, 2026 ruling?
A three-judge D.C. Circuit panel ruled 2-1 to uphold the Department of Defense’s designation of Anthropic as a supply-chain risk, with one judge dissenting, according to UPI and Politico.
Why did the Pentagon designate Anthropic a security risk?
The Department of Defense made the designation in March 2026. Anthropic’s attorneys argue it was retaliation for the company’s refusal to let its models be used for mass surveillance of Americans or lethal autonomous warfare, though the Pentagon’s full stated rationale has not been published in detail in available reporting.
What law is at the center of this case?
41 U.S.C. § 4713, a federal supply-chain-risk statute, is the provision the D.C. Circuit majority cited in upholding the Pentagon’s designation.
Did a court previously block the Pentagon’s designation?
Yes. The New York Times reported in March 2026 that a federal judge stayed the Pentagon’s labeling of Anthropic as a supply-chain risk, before the September 25, 2026 appeals ruling sided with the Department of Defense.
Does this affect Anthropic’s Claude models for non-government users?
Available reporting does not establish any restriction on Claude access for commercial or non-government customers tied to this case. The dispute centers on Defense Department procurement and use.
What happens if the D.C. Circuit denies the rehearing?
Anthropic’s reported options would likely include seeking review from the U.S. Supreme Court or continuing to contest the designation’s practical application, though available reporting does not confirm Anthropic’s specific next steps if rehearing is denied.
Could this affect other AI companies?
Legal observers and the Congressional Research Service have flagged the case as potentially precedent-setting for how federal supply-chain-risk authority could apply to other AI developers with similar usage-policy restrictions, though no other company has been named in this specific designation.
Related Coverage
- Chinese LLMs Risk $2.3T AI Capex Reset [2026]
- Anthropic Cuts AI Distillation Theft by Up to 99% [2026]
- Anthropic's AI Replication Warning: 0 Rivals Respond [2026]
- Anthropic Warns AI Could Clone Itself on $55 TVs [2026]
- Character.ai Bankruptcy: 163M Users, Rivals Brace [2026]
