Scoop: NSA nutzt Anthropics Mythos trotz schwarzer Liste
https://www.axios.com/2026/04/19/nsa-anthropic-mythos-pentagon
Scoop: NSA nutzt Anthropics Mythos trotz schwarzer Liste
https://www.axios.com/2026/04/19/nsa-anthropic-mythos-pentagon
19 Kommentare
Rules for thee, not for me.
Don’t worry everyone here says that Mythos is just hype and it can’t do anything that the previous models or open source can do.
Delusional people will tell you the NSA must be negotiating to use it, then leak that fact to prop up the bubble or something…
I’m not quite sure what sort of mental gymnastics are being done about the „anthropic is a supply chain risk“ desgination and how that fits into AI being a nothingburger, maybe they’ll just pull out the „it’s all FUD“ card and refuse to elaborate further.
They are hypocrites and liars? I’m shocked.
Probably running it against all the code they were able to exfil from targets or bar via NSL letters.
Does the NSA not know you can just get a fork off the leaked source code from github? Who the fuck are they even spying on?!
Islamaphobia and otherism from our govt led us to crafting the Patriot Act, and now the blowback effect has basically enabled all this spying and almost noone cares. Snowden should us a small sample and people called him a traitor. It doesn’t need to be said yet Ill say it, this spying network has almost never ever stopped a major or even a minor terrorist attack on home soil and sure is t being used to stop domestic terrorism. We are losing our privacy and paying for it with oue tax dollars and have ZERO say or ability to stop it.
After 9/11 almost everyone was fill charge Patriot Act, which led to DHS then to TSA and now ICE, its pathetic people are so capable of being gaslighted in surrendering our freedom and even more pathetic people dont care.
Lastly there are wildly strong and severe federal penalties for simply reading your neighbors mail, but its totally cool to read the same information just digitally AND to save/store it for an eternity.
A lot of agencies are using Anthropic models since the last injunction. This isn’t surprising. Even usai.gov is showing Anthropic’s logs since that ruling.
This regime is a POS.
Fuck the NSA.
The whole Anthropic thing is very much like the Iran war. It’s happening but at the same time it’s not happening. They are about to destroy Anthropic and at the same time they are having very good meetings.
If the government is using it it’s because Anthropic let them. Why would they do that?
I wish we had Mythos! We’re scraping the bottom of the barrel at my gov. They are still allowing Claude since the injunction. Hopefully we can still use it. A lot of AI is not certified for gov yet so you can only use it for public information.
Would have thought these alphabet agencies especially NSA would have had their own AI. Considering all the tools they have built to date.
✨️R e s t a r t your phone✨️
I’m sorry, but I’m not believing that DARPA and NSA don’t already have some wild AI shit we’ve never heard about.
This is an ad for Mythos.
I thought anthropic was blacklisted. Maybe I’m remembering wrong though.
It’s safe to assume Palantir has their own mythos product, as all the other ones either do or are working on one.
I’ll keep harping on this point. People jumped the gun on OpenAI while totally missing the point of using Local models.
OpenAI, Anthropic all have to generate revenue so they are going to sell their products to the government because it’s a guaranteed income.
If we want to be serious about offloading our AI usage then invest in local models or just not using AI at all.
Ok so why would the NSA not want access to Mythos, solely from the standpoint of testing what kind of threat is poses in the wrong hands?
Anthropic said themselves that the model’s ability to point out security exploits efficiently, along with how it tested in hacking games both should cause national security agencies to learn how and why… it’s kinda what they’re there for.
This has nothing to do with regime or who’s in charge and more about gauging a potential threat.