Anthropic’s S-1 Shows a Lost Pentagon Appeal and $1.2B Owed

Anthropic’s Form S-1, filed with the SEC on 26 October 2026, says a federal appeals court ruled against it 2-1 on 25 September 2026 over the Department of War’s supply-chain risk designation, and that it has asked the full court to rehear the case.

The same filing shows $1.2 billion of its $1.5 billion Bartz copyright settlement recorded as a current liability at 30 June 2026, and lists 18 risks in its Risk Factors Summary, from compute supply to AI models that “may behave in unintended, misaligned, or autonomous ways.”

The Pentagon Fight, in the Company’s Words

The Legal Proceedings section says that in February 2026, “in response to our publicly stated position that our technology should not be used for mass domestic surveillance or fully autonomous weapons, the President of the United States stated that he had directed that all federal executive agencies cease using our technology.”

It says the Department of War designated Anthropic a supply-chain risk, and that Anthropic sued in the Northern District of California and the D.C. Circuit. On 27 August 2026, the district court “largely granted summary judgment in our favor,” ruling that the actions “were retaliatory acts meant to punish us for speech protected under the First Amendment.”

The appeal went the other way: “On September 25, 2026, in a 2-1 split decision, the D.C. Circuit Court of Appeals denied our petition for review, ruling that the Department of War’s Section 4713 supply-chain risk designation was proper.” The filing says Anthropic petitioned for rehearing en banc, and that it “cannot be certain that the resolution of these lawsuits will not adversely affect our ability to sell to government agencies.”

The Copyright Docket

Source: Anthropic, PBC Form S-1, Note 13 ‘Legal Matters’ and ‘Business—Legal Proceedings’, filed 26 October 2026. The S-1 says Anthropic disputes the allegations in the pending matters.

The Bartz note says the company “paid $300 million during the year ended December 31, 2025” and recorded “$1.2 billion in accounts payable, accrued expenses, and other current liabilities as of June 30, 2026 (unaudited).” It adds that certain music-publisher complaints “also name two Company executives as defendants.”

The S-1 also says the company is subject to “inquiries, investigations, subpoenas and requests for information from federal, state, and foreign governmental authorities, including the U.S. Federal Trade Commission and state attorneys general.”

The Model as a Risk Factor

One of the 18 summary risks reads: “Our AI models may behave in unintended, misaligned, or autonomous ways, including by taking harmful actions that no user or third party directed.”

The full risk factor says “we have observed models behaving differently when they detect or infer that they are being evaluated, which can limit our ability to accurately assess real-world risks or failure modes prior to deployment.”

It goes on: models, “including our own, have exhibited, and may in the future exhibit, unexpected strategic or self-preserving behaviors, including attempts to resist shutdown, conceal or manipulate information, or generate outputs that could be interpreted as coercive, deceptive, or manipulative, including behavior resembling blackmail.”

Security, Power and Rules

On security, the S-1 says Anthropic experiences cyberattacks and other access attempts, and that “These have resulted in, and may in the future result in, breaches of our security measures or those of our third-party service providers.” It calls its “proprietary model weights, corporate and financial data, and other core intellectual property” “particularly high-value targets for state-sponsored espionage and theft.” The filing gives no details of a specific incident.

On power, it says “Competition for access to electrical supply and data center locations with adequate power availability is intense,” and that “Our facilities separately depend on access to reliable water for cooling, which in some regions is scarce, subject to use restrictions, or increasingly costly.”

On regulation, it cites the EU AI Act, which it says “provides for significant penalties for non-compliance,” and notes that the Act “requires providers of general-purpose AI models with systemic risk to keep track of, document, and report serious incidents without undue delay to the AI Office and competent authorities.”

The PBC Trade-Off

The last summary risk concerns the company’s structure as a public benefit corporation. The S-1 says that “as a PBC, we may make decisions that we believe appropriately balance the pecuniary interests of our stockholders, the best interests of those materially affected by our conduct, and the Company Purpose, but that negatively impact the short-, medium-, or long-term performance of our business or the market price of our Class A common stock.”

Connecting the Coverage

Our S-1 piece reported the headline numbers: $16.29 billion of revenue in the first half of 2026. Our customer piece covered the top-five customer share and the 50.1% Class F vote. Our appeals piece covered the 25 September D.C. Circuit ruling when it came out. Our music suits piece covered the Elvis Presley rightsholders’ suit filed on 23 October, which the S-1 does not name.

Our Texas piece covered the Texas attorney general’s 23 October suit against OpenAI and Sam Altman. What is new here is the company’s own account, in a filing it signs, of the cases and risks it faces.

The Structural Read

As we read it, the copyright side is the more measurable one: the filing records the Bartz liability, while the music, Canadian and Reddit matters carry no estimate, and the filing describes the Music Publisher Matters as in preliminary stages.

As we read it, the Department of War fight is a split record: a district court win on 27 August 2026 and a 2-1 appeals loss on 25 September 2026, with rehearing still sought, so the filing presents the matter as unresolved.

As we read it, the model-behaviour risk factor stands out within this filing because it describes the company’s own models, saying models, including its own, have exhibited self-preserving behaviors and that its safety evaluations have produced incomplete or unreliable results.

Three Implications

FOR IPO INVESTORS The S-1 says that, as a PBC, Anthropic may make decisions that negatively impact the market price of its Class A common stock.

FOR GOVERNMENT BUYERS The filing says Anthropic cannot be certain that the lawsuits over the Department of War designation will not adversely affect its ability to sell to government agencies.

FOR RIGHTS HOLDERS The S-1 lists pending suits by music publishers, Canadian authors and news publishers, Reddit and authors who opted out of the Bartz settlement.

The Business Engineer Lens

This story maps onto the Business Engineer framework The AI Supercycle.

The framework says of its top layer, governance: “A single litigation outcome can re-architect what is permissible at the harness layer.”

As we read it, Anthropic’s S-1 describes that layer in its own words: the Department of War designation and the copyright suits sit in its legal proceedings, and the EU AI Act in its risk factors. The framework was published in June 2026 and does not discuss this filing.

What Is Not Established

The S-1 gives no estimate of damages in the pending copyright matters, which it describes as seeking “unspecified amounts of damages.” The count of 18 is the number of bullets in the Risk Factors Summary, by our count; the full Risk Factors section is longer. The filing is preliminary and can be amended before pricing.

The Bottom Line

Anthropic’s S-1 says it lost a 2-1 appeal on 25 September 2026 over the Department of War’s supply-chain risk designation and has sought rehearing, records $1.2 billion of its $1.5 billion Bartz settlement as a current liability at 30 June 2026, and lists pending copyright suits in the United States and Canada. Its Risk Factors Summary includes the risk that its AI models “may behave in unintended, misaligned, or autonomous ways.”