Large-scale online deanonymization with LLMs

[Bootstrap]

It’s amazing to me how fine they draw a line.

To me, mass surveillance and killing people are not fine lines.

[Bootstrap]

I believe in civilian control of the government and the military; that includes AI tools, and some unaccountable private contractor should not be in charge of what the military can or cannot do.

The proper way to regulate how the government and military use LLM tools is via Congress and the democratic process, not a private third party.

I also believe in civilian control of the military. Here's what it means: the armed forces are subordinate to elected civilian leadership. It protects democracy from military takeovers.

Civilian control of the military does not give the military control over the private sector or civilian society. If that happens, you’re no longer talking about civilian control of the military, but about military control of civilians.

[JohnH]

I also believe in civilian control of the military. Here's what it means: the armed forces are subordinate to elected civilian leadership. It protects democracy from military takeovers.

Civilian control of the military does not give the military control over the private sector or civilian society. If that happens, you’re no longer talking about civilian control of the military, but about military control of civilians.

Right, but civilian control of the military does mean the military can decide not to do business with a private entity.

Note that Anthropic is already suing the military to apparently force them to buy their defective products which the military doesn't want.

And Congress long ago passed laws that allow the U.S. government to require civilian manufacturers to provide products when there is a clear military / national interest need. If you don't like that, you need to repeal the law. It is entirely constitutional.

[Bootstrap]

I also believe in civilian control of the military. Here's what it means: the armed forces are subordinate to elected civilian leadership. It protects democracy from military takeovers.

Civilian control of the military does not give the military control over the private sector or civilian society. If that happens, you’re no longer talking about civilian control of the military, but about military control of civilians.

Right, but civilian control of the military does mean the military can decide not to do business with a private entity.

Note that Anthropic is already suing the military to apparently force them to buy their defective products which the military doesn't want.

And Congress long ago passed laws that allow the U.S. government to require civilian manufacturers to provide products when there is a clear military / national interest need. If you don't like that, you need to repeal the law. It is entirely constitutional.

Did Anthropic file a lawsuit to force the military to buy its products? I haven’t seen any reporting of that. Do you have a source?

I did see that they said they would challenge a “supply-chain risk” designation in court. That designation basically means the government has decided the company is too risky to do business with and can block it from federal contracts.

If they win the lawsuit, it wouldn’t force the government to buy their software. It would just mean they’re allowed to compete for contracts again.

You called the products “defective,” but the issue seems to be that the DoD wants access to them. Those two things don’t really go together - why would the DoD try to force them to use them for these purposes if they think they are defective? From what I’ve read, Anthropic’s models are considered some of the top LLMs right now. If there’s a source saying the DoD thinks they’re defective in any objective, technical sense, I’d like to see it.

[ken_sylvania]

Worth noting: the government is suing Anthropic because they do not allow their LLMs to be used for two specific things:

1. Mass surveillance
2. Making the decision whether an autonomous weapon should kill someone without prior human approval

I don't think LLMs should be used for either purpose. It alarms me that the government is publicly suing Anthropic over this, and even threatening to use wartime production powers if Anthropic does not back down.

Also alarming: Anthropic is alone in this. OpenAI, Gemini, and Grok do not have restrictions like these.

I hadn't heard about that and I can't find any information about it online. Care to share your source for this?

Google this:

Anthropic government lawsuit mass surveillance autonomous weapons

Or this:

site:reuters.com Anthropic lawsuit Pentagon

This is also useful:

https://www.theverge.com/ai-artificial- ... n-soul-doc

Neither option brings up anything about the government suing Anthropic. Various references to the fact that Anthropic might sue the government but nothing saying the government has already sued Anthropic.

[JohnH]

Did Anthropic file a lawsuit to force the military to buy its products? I haven’t seen any reporting of that. Do you have a source?

https://www.axios.com/2026/02/28/anthro ... ai-dispute

I did see that they said they would challenge a “supply-chain risk” designation in court. That designation basically means the government has decided the company is too risky to do business with and can block it from federal contracts.

Right. That's a lawsuit. In other words, they want to force the government to do business with them.

If they win the lawsuit, it wouldn’t force the government to buy their software. It would just mean they’re allowed to compete for contracts again.

Pray tell, what is this distinction?

You called the products “defective,” but the issue seems to be that the DoD wants access to them. Those two things don’t really go together - why would the DoD try to force them to use them for these purposes if they think they are defective? From what I’ve read, Anthropic’s models are considered some of the top LLMs right now. If there’s a source saying the DoD thinks they’re defective in any objective, technical sense, I’d like to see it.

The issue at hand is that the DoD wants to be able to use AI technology for "any lawful purpose".

Anthropic apparently thinks they are a shadow government who can write their own "constitution" and tell the government what to do. That is a quite defective product, particularly in the area of AI technology.

Anthropic is free to go find non-government customers who are willing to submit to their "constitution".

[Bootstrap]

Did Anthropic file a lawsuit to force the military to buy its products? I haven’t seen any reporting of that. Do you have a source?

https://www.axios.com/2026/02/28/anthro ... ai-dispute

I did see that they said they would challenge a “supply-chain risk” designation in court. That designation basically means the government has decided the company is too risky to do business with and can block it from federal contracts.

Right. That's a lawsuit. In other words, they want to force the government to do business with them.

I don't think so.

If they win the lawsuit, it wouldn’t force the government to buy their software. It would just mean they’re allowed to compete for contracts again.

Pray tell, what is this distinction?

Whenever they compete for a contract, the agency decides who they want to buy from, and there are usually multiple vendors competing for the contract.

The government is never forced to choose Anthropic. But if the agency believes Anthropic has the best software for their needs, they are allowed to choose it.

[JohnH]

The DoD has assessed Anthropic to be an unacceptable risk.

Anthropic’s answer is a lawsuit.