pull down to refresh

An AI company that is against mass surveillance. Yes, really.

Archived link: https://archive.is/IBvWt

Defense Secretary Pete Hegseth is "close" to cutting business ties with Anthropic and designating the AI company a "supply chain risk" — meaning anyone who wants to do business with the U.S. military has to cut ties with the company, a senior Pentagon official told Axios.
The senior official said: "It will be an enormous pain in the ass to disentangle, and we are going to make sure they pay a price for forcing our hand like this."
Why it matters: That kind of penalty is usually reserved for foreign adversaries.
Anthropic's Claude is the only AI model currently available in the military's classified systems, and is the world leader for many business applications. Pentagon officials heartily praise Claude's capabilities.
As a sign of how embedded the software already is within the military, Claude was used during the Maduro raid in January, as Axios reported [1] on Friday.
Anthropic CEO Dario Amodei takes these issues very seriously, but is a pragmatist. Anthropic is prepared to loosen its current terms of use, but wants to ensure its tools aren't used to spy on Americans en masse, or to develop weapons that fire with no human involvement.
The Pentagon claims that's unduly restrictive, and that there are all sorts of gray areas that would make it unworkable to operate on such terms. Pentagon officials are insisting in negotiations with Anthropic and three other big AI labs — OpenAI, Google and xAI — that the military be able to use their tools for "all lawful purposes."
A source familiar with the dynamics said senior defense officials have been frustrated with Anthropic for some time, and embraced the opportunity to pick a public fight.

The Pentagon knows from experience that they cannot comply with a "no mass surveillance of Americans" requirement if they intend to build a dragnet, because at scale you not want to have to censor away data - that data can be relevant to your pursuits, whatever those may be. If you need dragnet surveillance, you are going to have to be a dragnet surveillance maxi, or be "a loser". It's not worth further eroding public trust if you're going to not be able to use your data against them.

We can be quite certain that dragnets are exactly what is being built, across multiple lateral projects and what looks like cross-layering, from intel like #1436017, and it is proven that OpenAI is actively participating in this, not just as a customer, but also as a supplier to Palantir.

Thus, either Anthropic budges, or they're out. It will be interesting to see what they choose.

  1. archived: Pentagon's use of Claude during Maduro raid sparks Anthropic feud (Feb 13, 2026)

142 sats \ 5 replies \ @OT 11h

Sounds like they have some basic principles. I wonder how long they can hold out.

It's pretty frightening really. The most powerful military teaming up with a super intelligence...

reply

It's not a super intelligence... yet. Right now really just another LLM but one that is pretty well trained at safety, tool execution and instruction following. And research (+ my own "play" experience) suggests that the bot instructing the bot still has limits in creativity and really only works well in pre-trained scenarios (#1437544)

So I'd pose that it's still more of a pattern matching machine than an actual intelligence. It doesn't ask why? when it is not trained to ask that, though you can simulate that by adding an instruction to always ask why. The problem is that it doesn't ask why it has to ask why then.

reply
121 sats \ 3 replies \ @OT 10h

Consider doing an AMA. I got lots of questions

reply

AMA anytime (AMAA 😂). I always bookmark honest questions and work through them when off work. Just put in the list!

reply
121 sats \ 1 reply \ @OT 9h

OK then...

It's a bit out place here but I've been getting some serious fomo on setting up Openclaw. I think most ideas I have are more creative and so I don't need to be paranoid on security/privacy. The main idea I keep obsessing over is setting up a bunch of bots trained on specific instruments. I want them to be able to play a track or come up with an idea. It might be too complex to start with but I'd be interested to hear if you have any tips for getting this up.

Also you mentioned in an earlier post that you have everything Openclaw can do already. It is a distraction? Or a good place to start?

reply

Answering in re-arranged order.

I think most ideas I have are more creative and so I don't need to be paranoid on security/privacy.
[..]
Also you mentioned in an earlier post that you have everything Openclaw can do already. It is a distraction? Or a good place to start?

It's a bad place to start for anyone that isn't an IT security / opsec UltraChad because it by default offers zero security against anything; i.e. you install it, you're open for anyone in the world to leech all your data that you yourself have access to (and even things you forgot or never knew you had access to).

If however you know within a few seconds without asking an LLM how you can improve upon my setup for claude code to make it less risky, because there simply are things missing in that comment that I forgot mentioning and things I learned since, then you may be that. Otherwise, you're going to be at-risk from the first time you run it.

Don't take my word for it, there's a nice article about this posted at #1437998

But:

  • Anyone can get help from a friend that does know what they're doing (just make sure that they really know what they're doing) and reduce the risk significantly
  • I have it on my list to check out nullclaw, but I think that because the concept/architecture itself is risky-by-design, it will be hard to have normal usage. However, that one may be doable with proper instructions - I will tag you when I find one that works.

That said,

The main idea I keep obsessing over is setting up a bunch of bots trained on specific instruments. I want them to be able to play a track or come up with an idea. It might be too complex to start with but I'd be interested to hear if you have any tips for getting this up.

I think that technically you don't need "a bunch", but just one that can interface with MIDI or a tracker (or both!). I haven't touched MIDI in at least a decade so I'm going to stick with a tracker and probably the most modern one right now is strudel.cc. It also has great docs that you can probably let an LLM run through.

Depending on what subscriptions you have in place, you may not need any claw stuff for this. Both GPT and Claude have co-working / rich interface modes, and you could probably even prompt a script from a text/coding chatbot if you really wanted and feed the output into a repl like strudel.

The real challenge for your project? Learning how successful humans compose and produce music and finding creative ways to automate the steps.

reply
21 sats \ 0 replies \ @Solomonsatoshi 13h -100 sats

Why don't you attach and show LN wallets?

Don't you want to use LN and sats as much as possible in order to support the development of the LN?

Because that's what attaching and showing attached wallets optimizes.

Nearly everyone else who uses this platform regularly and is a serious Bitcoiner has already attached and shown their LN wallet- but not you- Why?

Are you just a Big talk and No Walk Hypocrit?

21 sats \ 0 replies \ @Solomonsatoshi 13h -200 sats

Attaching and showing attached wallets is both maximising your use of sats and LN and shows all other users that you are serious about using Bitcoin and supporting the growth and strength of the LN.

The vast majority of regular users here have already attached and show they have attached wallets.

It is not difficult but it is something most of us can do that will have a positive and constructive effect.

Difficult to understand anyone describing themself as a Bitcoiner who cannot be bothered to attach and show attached LN wallets.