llms.blog
Anthropic

Judge says record has gotten worse for government in Anthropic dispute

Judge says record "has gotten worse" for government in Anthropic dispute A federal judge appears poised to again side with Anthropic in its legal battle against the Trump administration's governmentwide ban on the company's technology, stating during a Thursday hearing that the government's case has deteriorated since she issued a preliminary injunction in March. Judge Rita Lin of the U.S. District Court for the Northern District of California presided over competing motions for summary judg

axc
axc
3 min
Judge says record has gotten worse for government in Anthropic dispute

Judge says record "has gotten worse" for government in Anthropic dispute

A federal judge appears poised to again side with Anthropic in its legal battle against the Trump administration's governmentwide ban on the company's technology, stating during a Thursday hearing that the government's case has deteriorated since she issued a preliminary injunction in March.

Judge Rita Lin of the U.S. District Court for the Northern District of California presided over competing motions for summary judgment in Anthropic PBC v. Department of War, the case stemming from the administration's decision to designate Anthropic as a supply-chain risk and ban federal use of Claude.

The hearing

Lin said the record in the dispute appears mostly the same as what she reviewed before issuing her March injunction, which blocked the government from enforcing the ban and halted the Department of Defense's supply-chain risk designation.

"If anything, it seems like the record, in some ways, has gotten worse for the government," Lin said.

The record included a batch of emails between Anthropic CEO Dario Amodei and DOD research and engineering leader Emil Michael, released as part of the litigation. Lin said she found no evidence that Anthropic could alter the model after providing access to the government, that it possessed a "kill switch," or that the model was delivered with nefarious controls.

Rather, Lin said, the record suggests that Anthropic publicly criticized the DOD, and that criticism led to the government's stated lack of trust in the AI provider. She characterized the government's position as arguing that if a defense contractor criticizes the administration, that criticism can itself amount to a breach of trust justifying retaliatory action, including contract termination and labeling the company an enemy of the state.

"If that's really what the government's position is," Lin said she finds it "to be really troubling."

Current status of Claude in government

James Harlow, a Justice Department lawyer representing the government, told the court that the DOD continues phasing out Claude systems with a completion date of September 30. Another unnamed defendant agency had preexisting plans to transition away from Anthropic to a different AI provider.

Harlow noted that elsewhere in government, some agencies are still experimenting with Claude through sandboxes, pilot programs, and third-party vendors. He referenced a pilot program set to end August 30, almost certainly the General Services Administration's OneGov deals, through which several AI providers have offered products at discounted rates.

Michael Mongan, a lawyer with Wilmer Hale representing Anthropic, said two unnamed agencies paused negotiations on new contracts, while other federal agencies continue pursuing new deals. Anthropic is not aware of agencies canceling contracts since the preliminary injunction, though a contract with the Pentagon's Chief Digital and Artificial Intelligence Office was canceled before the court granted injunctive relief.

The Mythos question

A separate dispute emerged in June when the Trump administration restricted Anthropic's Mythos 5 and Fable 5 systems under export controls, citing national security concerns. Those restrictions were lifted weeks later after the administration and Anthropic agreed on guardrails.

Lin asked the government to provide details about expanded use of Anthropic tools in national security, including deployment of Mythos. She noted that expanded use of Anthropic's services would be inconsistent with DOD's arguments in the case. Reports from Reuters and Axios have suggested that government agencies, including the NSA, are using Mythos.

Harlow initially argued the question solicited nonpublic information and was irrelevant, but later indicated the government may be able to submit something in response.

Background

The dispute began in early 2026 when Anthropic went public with details of its disagreement with the Pentagon over how its technology was being used. In response, President Donald Trump and Defense Secretary Pete Hegseth issued directives via social media banning federal agency use of Anthropic technology and announcing the supply-chain risk designation.

Those actions had immediate impact, with agencies across the government halting internal use of Claude. Anthropic challenged the actions in both district and circuit court. Lin's March preliminary injunction temporarily allowed agencies to continue using Anthropic tools while the case proceeds.

Sources

AnthropicPolicyLegalGovernment
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Decentralized and Peer-to-Peer LLM Inference in Production: Architecture, Ring Memory Partitioning, and Network Latency

    Decentralized and Peer-to-Peer LLM Inference in Production: Architecture, Ring Memory Partitioning, and Network Latency Frontier open-weight models such as Llama 3.1 405B, DeepSeek-V3, and Command R+ have expanded model capabilities, but their parameter scales exceed the physical memory limits of individual consumer and edge workstations. Running a 405-billion parameter model in 16-bit precision requires over 810 GB of memory, and even 4-bit quantized variants require roughly 230 GB of contiguo

    1 min

  • Discrete Diffusion in Large Language Models: How Continuous-Time Markov Chains, Absorbing States, and Score Entropy Challenge Autoregressive Generation

    The dominance of autoregressive architectures in large language models rests on a fundamental mathematical formulation: the chain rule of probability. By factoring the joint distribution of a sequence into a product of conditional probabilities, $p(x) = \prod_{i=1}^N p(x_i \mid x_{<i})$, autoregressive models reduce text generation to sequential next-token prediction. While this left-to-right causal factorization has scaled effectively across compute regimes, it imposes rigid operational constr

    1 min

  • AI Agents Surpass Humans on OpenRouter as Agentic Token Usage Jumps 14x

    Autonomous AI agents have overtaken human users as the primary consumers of language model compute on OpenRouter, with agentic token volume surging fourteenfold over the past six months. Data published by OpenRouter analyst Peter Walker indicates that February 6 marked the permanent inflection point where token consumption by automated agents exceeded direct human API traffic. Since that threshold, agentic token volume on the multi-model gateway has climbed from 0.51 trillion to 7.3 trillion to

    1 min