Anthropic Wins Court Order but Faces No Pentagon Procurement Guarantee
Bloomberg’s Rachel Metz reports that a California judge has ordered the Trump administration to lift its supply-chain-risk designation against Anthropic, finding it inadequately justified and partly intended to make an example of the AI company. The ruling does not require the Pentagon to use Anthropic’s products, Metz says, and a separate case in Washington remains unresolved. Separately, Anthropic, OpenAI and more than 100 organizations are warning that increasingly capable AI models could make cyberattacks more widespread even as they strengthen cyber defense.
Anthropic’s court win removes a designation, not the Pentagon’s discretion
A California judge has ordered the Trump administration to lift its ban on Anthropic’s AI technology for federal agencies. The ruling found that the supply-chain-risk designation was not adequately justified and was based partly on a desire to make a public example of Anthropic for criticizing the government.
That result answers one question: the challenged designation cannot remain in place under the California order. It does not answer two others that matter for Anthropic’s federal business: whether the Defense Department must buy or use its products, and whether the government can still prevail in a separate Washington, D.C., case.
Rachel Metz described the ruling as a positive decision for Anthropic, but not a final resolution. The company and the government have been fighting on two legal fronts. Anthropic won in California; the D.C. proceeding remains outstanding, and Metz said its outcome will shape what happens next.
The immediate procurement consequence is more limited than a restoration of Anthropic as a required government supplier. The judge’s order does not compel the Defense Department to use Anthropic technology. Metz said the department remains free to take lawful actions that were available before it imposed the supply-chain-risk designation, provided those actions comply with applicable regulations, statutes, and constitutional provisions.
“The order that the judge made the other day, it doesn't mean that the Defense Department has to use Anthropic's products,” Metz said. “It could go and use another supplier.”
That distinction separates the legal defeat from the procurement decision. The administration must lift the challenged label, but the Pentagon can still select a different AI provider. Metz noted that the department already works with other companies, including OpenAI. For Anthropic, the ruling removes one government-imposed obstacle without creating a guaranteed contract or deployment.
The underlying conflict is over limits on military and security uses
The dispute arose from Anthropic’s effort to limit how its technology could be used. Tim Stenovec identified two uses the company did not want to support: mass surveillance of Americans and autonomous weapons deployment.
The government resisted those restrictions on use because of national-security concerns. The California decision addressed the supply-chain-risk designation imposed on Anthropic, not a broader settlement of that disagreement over acceptable uses of AI in military and security contexts.
That leaves the central tension intact. Anthropic sought limits on applications it considered unacceptable; the government objected to those limits on national-security grounds. The legal order removes the designation at issue while the separate D.C. case remains unresolved.
Cyber-defense capabilities can also be used to attack
Anthropic’s legal dispute sits alongside a separate warning from AI, technology, and financial-services companies: systems that can help defenders respond to cyber threats may also make offensive cyber activity easier for bad actors.
An on-screen statement attributed to “A call for collective action on cyber defense” warned: “In the coming months, AI-enabled cyber attacks will become far more widespread and sophisticated as models around the world become increasingly capable.” A second graphic listed signatories including OpenAI, Anthropic, Accenture, Cisco, Figma, Google, IBM, and Microsoft.
Rachel Metz said there had already been a handful of incidents in recent months involving cyber-capable models from OpenAI and Anthropic, along with an incident involving one of Meta’s models. She did not provide details of those incidents.
The companies’ concern is the dual use of the systems. Metz said the models are highly capable of defending against cybersecurity threats, but can also be used “for the opposite,” to perpetrate those threats. The signatories are urging governments, organizations, and companies to shore up their existing defenses as models improve.
The two issues concern different risks, but each centers on how increasingly capable AI systems are used. In the federal dispute, the live questions include supplier discretion and limits on military or security applications. In the cyber warning, the companies are asking institutions to strengthen defenses against misuse by attackers.