We do not have a "rubylik" from Claude models in Pentagon closed systems.

Anthropic, the developer of the Claude AI model, stated that after deploying its models in closed Pentagon systems it has no access to those systems and cannot influence their operation. This statement refutes the claims of Defense Secretary Peter Hegseth, who asserted that Anthropic poses a threat to the supply chain.

1. Technical deployment details
* Case No. 26‑1049 – a 60-page lawsuit filed in the U.S. Court of Appeals for the District of Columbia Circuit.
* The complaint describes Claude’s architecture within the Pentagon: the model is “static” – it does not change on its own and does not degrade.
* Anthropic cannot:
* Send updates;
* Introduce vulnerabilities;
* Disable the system remotely;
* Have a backdoor or remote access to Department of Defense (DoD) infrastructure.
* Before deployment, the government independently tests each model and conducts multi‑layered analysis of its behavior.
* If Anthropic engineers gain clearance to classified data, they can work with the model only at the agency’s discretion and under strict oversight.

2. Contradiction of the minister’s arguments
Secretary Hegseth and Research Deputy Emile Michael claimed that Anthropic seeks an “operational veto” over military operations and could secretly alter Claude or disable it in combat.
Anthropic rejects these claims as unfounded, because it physically cannot perform any of the listed actions.

3. Claude usage policy
* The dispute arose over two policy points regarding permissible use:
* Prohibition on using Claude for lethal autonomous weapons;
* Prohibition on mass surveillance of U.S. citizens.
* Anthropic believes the model is not sufficiently reliable for these tasks: an error in an autonomous combat decision could cost lives of service members or civilians, and existing legal frameworks do not account for the scale of data processing by AI systems.

4. Negotiations with the Pentagon
* In September 2025, the Pentagon first demanded a “full lawful use” approval for Claude.
* After several months of talks:
* On February 24 2026, Hegseth met personally with Anthropic CEO and co‑founder Dario Amodei, praised the model’s capabilities, and proposed cooperation.
* At the same time he issued an ultimatum: if the company does not agree by February 27, it will be declared a supply‑chain threat.
* On February 26, Amodei publicly replied, emphasizing that decisions on military operations are made by the Pentagon and that Anthropic “cannot in good conscience” accept the minister’s terms. He highlighted the contradiction of simultaneously labeling the company a threat while demanding its presence in combat systems.

5. Business consequences
* The threat status extends to all Anthropic products and services, as well as any direct or subcontracted contracts with the Pentagon.
* DoD terminated a two‑year agreement worth up to $200 million.
* A memorandum dated March 6 required all agency units to cease using Anthropic products and remove the company from approved vendor lists within 180 days.

6. Litigation
* The U.S. Court of Appeals for the District of Columbia denied Anthropic an injunction against the threat status but scheduled expedited substantive review.
* The court acknowledged that the company presents serious arguments and could suffer irreparable harm during proceedings.
* New hearings are set for May 19 2026.

Conclusion:
Anthropic asserts it lacks the technical ability to influence its models in the Pentagon, thereby refuting Secretary Hegseth’s accusations. Despite this, the company faces significant legal and commercial repercussions due to the government’s supply‑chain threat designation. The litigation will continue over the coming months.