First Amendment retaliation. Anthropic alleges that Pentagon officials illegally retaliated against the company for its position on AI safety. They argue that Trump, Hegseth, and others wanted to punish Anthropic for protected speech, citing public social media and other dialogue as evidence that the punishment is ideological in nature.
Misuse of the supply chain risk designation. Anthropic was officially designated a supply chain risk, which requires defense contractors to certify they don’t use Claude in their Pentagon work. Anthropic argues that this is a misuse of the SCR designation which Congress intended to stop foreign actors, and that Anthropic clearly does not pose a supply-chain risk in a reading of the law.
Lack of Due Process (Fifth Amendment violation). “The Challenged Actions arbitrarily deprive Anthropic of those interests without any process, much less due process.”
Ultra vires. Anthropic alleges that the Presidential Directive requiring every federal agency to immediately cease all use of Anthropic’s technology exceeds the limits of the President’s authority as granted by Congress.
Administrative Procedural Act. Similar to the above, Anthropic argues that various agencies in the administration violates the administrative procedural act, and the sanctions are not permitted to the relevant agencies as a duty granted to them by Congress.
IANAL etc. In my personal opinion, #2 seems very clearcut from a common-language and precedent reading of these things[1]. #1 also seems strong. Sources I randomly skimmed online thought Anthropic had a good case in #3-#5 too, but I don’t have an independent view (I don’t immediately see a strong in-principle reason why the executive branch can’t do these things, so the decision probably rests on a combination of specific wordings of past legislation + case law that I don’t have a strong reason or need as a non-expert to understand).
As expected, Anthropic cites the recent precedent of NRA v. Vullo (for #1). #5 is also a strong case because while the executive branch has pretty broad discretion on a number of policy decisions, the law requires a lot of due process and deliberation, and Anthropic argues that even if the designation is within the powers granted under the APA, the process was also “arbitrary and capricious” (see pages 7⁄8 in the complaint).
In my non-expert opinion, the first amendment retaliation case here is pretty strong, especially considering existing precedent.
On the off chance anybody is both interested in AI news and missed it, Anthropic sued the Department of War and other government officials/agencies for the supply chain risk designation in DC and Northern Californian Circuit. The full-text of the Northern Californian complaint here:
The primary complaints:
First Amendment retaliation. Anthropic alleges that Pentagon officials illegally retaliated against the company for its position on AI safety. They argue that Trump, Hegseth, and others wanted to punish Anthropic for protected speech, citing public social media and other dialogue as evidence that the punishment is ideological in nature.
Misuse of the supply chain risk designation. Anthropic was officially designated a supply chain risk, which requires defense contractors to certify they don’t use Claude in their Pentagon work. Anthropic argues that this is a misuse of the SCR designation which Congress intended to stop foreign actors, and that Anthropic clearly does not pose a supply-chain risk in a reading of the law.
Lack of Due Process (Fifth Amendment violation). “The Challenged Actions arbitrarily deprive Anthropic of those interests without any process, much less due process.”
Ultra vires. Anthropic alleges that the Presidential Directive requiring every federal agency to immediately cease all use of Anthropic’s technology exceeds the limits of the President’s authority as granted by Congress.
Administrative Procedural Act. Similar to the above, Anthropic argues that various agencies in the administration violates the administrative procedural act, and the sanctions are not permitted to the relevant agencies as a duty granted to them by Congress.
IANAL etc. In my personal opinion, #2 seems very clearcut from a common-language and precedent reading of these things[1]. #1 also seems strong. Sources I randomly skimmed online thought Anthropic had a good case in #3-#5 too, but I don’t have an independent view (I don’t immediately see a strong in-principle reason why the executive branch can’t do these things, so the decision probably rests on a combination of specific wordings of past legislation + case law that I don’t have a strong reason or need as a non-expert to understand).
The DC complaint looks less meaty (and I didn’t read it)
And my general prior is that common-language readings of law, while imperfect, is pretty good.
As expected, Anthropic cites the recent precedent of NRA v. Vullo (for #1). #5 is also a strong case because while the executive branch has pretty broad discretion on a number of policy decisions, the law requires a lot of due process and deliberation, and Anthropic argues that even if the designation is within the powers granted under the APA, the process was also “arbitrary and capricious” (see pages 7⁄8 in the complaint).
In my non-expert opinion, the first amendment retaliation case here is pretty strong, especially considering existing precedent.