Hegseth to Anthropic: Nice company you got there…

[Axtremus]

Don't think there is a meaningful distinction between different "versions" of a large AI model. It costs way too much for a company to perform parallel trainings to develop two significantly different "versions."

If the underlying AI model is the same, then putting different "guardrails" around the model gives little confidence that a tech savvy user (like the Pentagon) won't have the capability to get around those guardrails.

It sounds to me Anthropic wants the Pentagon to promise (as a matter of contract) not to use its product for certain purposes, rather than trying to not provide a certain "version" of its product.

[Axtremus]

https://www.washingtonpost.com/opinions/2026/02/26/hegseth-anthropic-ai-model-claude/

Hegseth wants Anthropic to modify its contract to allow “any lawful use” of the technology. Anthropic is willing to rewrite its current terms of use but not to include mass surveillance of Americans or accommodate weapons that operate without a person in the loop to make the final decision.

It seems WaPo's Editorial Board reads the situation as I do: it's a contractual "terms of use" issue, not a "product version" issue.

[89th]

https://www.anthropic.com/news/statement-department-of-war

[Horace]

OpenAI is in negotiations with the DoD to take over for Anthropic. I hope the "supply chain risk" threat is just negotiation hot air. I suspect it will prove to be, but who knows.

In the letter, Anthropic acknowledges that the law is currently ambiguous to non-existent. Their concerns are ethical rather than (currently) legal, and I think the DoD is well within its duties to find a vendor who will not impose their own ethical constraints.

[jon-nyc]

[jon-nyc]

Now it’s the Defense Production Act.

Hegseth can’t decide if Anthropic is a risk to national security or essential to national security.

[mark]

We have very dangerous idiots running our country.

[jon-nyc]

[Axtremus]

Srsly, the alternative would have done better than this.

[Renauda]

@Axtremus said in Hegseth to Anthropic: Nice company you got there…:

Srsly, the alternative would have done better than this.

As would the county dog catcher.

[xenon]

Unmitigated idiocy at the helm.

[Horace]

The threat to classify Anthropic as a supply chain threat is idiotic. The conflicting motivations of a company that brands itself with the highest ethical standards, but who want to provide cutting edge technology to the DoD, seems reasonable and even inevitable.

[Axtremus]

https://www.nbcnews.com/tech/tech-news/trump-bans-anthropic-government-use-rcna261055

... “OpenAI CEO Sam Altman posted on X Friday night that the company had struck a deal with the Department of Defense to deploy its models on the department’s classified networks.”

[jon-nyc]

[Horace]

OpenAI goes where Anthropic dared not to:

Note that OpenAI's spin of the agreement makes it sound identical to what Anthropic demanded and the DoD refused. I wonder what the practical differences are.

[Axtremus]

[Horace]

@Axtremus my post contained the crucial insight that openAI’s agreement is described as identical to what Anthropic offered. Therefore my contribution is objectively superior to yours.

[Mik]

My take? It is a discussion best approached quietly, not with the usual public belligerence of this administration. It looks petulant whether they are right or wrong and Anthropic looks like good citizens. The company framed this beautifully. Sometimes Trump et al just don't seem to know what fights are not wise to pick.

[jon-nyc]

Seems to me the important part happened after that. The remedy if they don’t give you what you want is to find a different vendor. Not crush what very well may be the leading AI company in the country.

[Mik]

Yeah, that too. If I were a CEO it would not make me eager to deal with this administration, preferring to see who is next. But that kind of ties in with what fight to pick.