Anthropic told a U.S. appeals court that it cannot remotely alter or “manipulate” its Claude artificial intelligence system once the tool is installed in classified Pentagon military networks, a position the company said is meant to counter the Trump administration’s argument that the rapidly expanding AI vendor poses a national-security “supply chain” risk.
The company made that argument in a 96-page filing submitted to the U.S. Court of Appeals in Washington, D.C., where a panel is scheduled to hear oral arguments on May 19. The filing comes as part of an ongoing legal fight that began after a contract dispute involving how AI technology can be used in fully autonomous weapons and the potential for surveillance of Americans.
Anthropic’s filing is aimed at responding to what the company described as the Pentagon’s effort to brand it with a designation intended to help prevent sabotage of national security systems by foreign adversaries. Anthropic told the appeals court that it should not be treated as a controllable supply-chain vulnerability, arguing instead that the Pentagon’s actions amount to illegal retaliation for the contract disagreement.
In the same filing, Anthropic said an earlier setback in the Washington case followed the appeals court’s rejection of the company’s request for a court order that would have paused the Pentagon’s actions while the panel gathers evidence. Anthropic said the new submission is intended to address some of the questions the appeals court has raised ahead of the scheduled arguments.
The company noted that it has already gained traction in a separate case in federal court in San Francisco that focused on similar issues. Court filings cited by the appeals-court submission say that after Anthropic prevailed in that other matter, the Trump administration removed the stigmatizing labels from Anthropic, but that a comparable order had not issued in the Washington case.
Anthropic described that difference as leaving it exposed to continued legal and operational uncertainty in the appeals-court fight. The company’s tools, it said, have helped position Anthropic as a growing technology presence alongside rival OpenAI, even as the Pentagon canceled a $200 million contract with Anthropic after the companies disputed the terms of how AI would be used.
In the aftermath of the cancellation, the Pentagon turned to OpenAI to provide its technology to the U.S. military, according to the filing summarized by the Associated Press. The appeals-court case now turns on whether the government can characterize Anthropic as a controllable risk at the point of classified deployment, and whether Anthropic’s allegations about retaliation are supported as the court proceedings continue.
The appeals-court panel will consider the Trump administration’s response, which the government is expected to file before the May 19 hearing.