logoalt Hacker News

NSA lost access to Mythos amid Anthropic dispute

203 pointsby thmtoday at 11:45 AM184 commentsview on HN

Unlocked: https://www.nytimes.com/2026/06/23/us/politics/nsa-lost-acce...


Comments

sometimelurkertoday at 11:02 PM

I really seriously doubt this. no proof, obviously, but this sounds too unbelievable. the NSA probably has the mythos pretrain and might be finetuning it themselves.

teravortoday at 7:49 PM

mythos allowed mediocre people to get results by holding their hand through the process, or just ignoring their irrelevant input and knowing what to do.

if you throw millions of tokens at IDA Pro MCP with the right prompt lets just say security by obscurity fails miserably because there is no obscurity when the LLM chews through the decompilation.

show 3 replies
gaiagraphiatoday at 10:23 PM

Is there a historical precedent as to what happened when the upstart denied capability to the empire?

The closest I can think of is the bronze age collapse.

show 3 replies
chasiltoday at 9:58 PM

'Mythos “broke into almost all of our classified systems, not in weeks, but in hours.”'

Is Mythos a significant danger?

The curl experience does not suggest that hysteria is warranted, but this gives me pause.

show 3 replies
zb3today at 8:34 PM

> That contract has not been finalized, and some Pentagon officials want the N.S.A. to find a way to work with other models.

Good, fsck NSA, that's the last organization I'd ever want to have access to Mythos. I hope this administration's incompetence will prevent them from regaining access for as long as possible

show 1 reply
medlaziktoday at 6:33 PM

AI marketing bullshit stunts are unlike anything I've seen in 30 years. It started with MS Copilot so called capabilities for work, which were completely made up use cases that didn't work at all (3 years later still). We've had OpenAI "AGI is coming" and "AI will take your job", now we have Mythos being so "dangerous" for cybersecurity, which of course makes the average Joe interpret it as Anthropic being "the better overall company, the NSA uses it!!". I mean gov foes with Anthropic are probably true, but the marketing is to blame not Mythos capabilities. This is all so fucking pathetic

show 7 replies
AustinDevtoday at 6:25 PM

They could easily take the weights if they wanted. I don't believe they meaningfully lost access.

show 6 replies
ck2today at 6:29 PM

they are doing DOGE-cuts to all of intelligence now anyway

dozens upon dozens fired for no reason

so US "intelligence" is going to go even further backwards

* https://www.yahoo.com/news/politics/articles/trump-acting-ch...

November is going to be insanity

show 5 replies
SadErntoday at 10:46 PM

[dead]

micromacrofoottoday at 6:32 PM

[flagged]

show 1 reply
charcircuittoday at 7:12 PM

Everyone lost access. What even is mainstream news these days.

show 1 reply
Madmallardtoday at 5:57 PM

Doesn't make any sense. They could just force them to provide Mythos to the federal government.

show 6 replies
gsibbletoday at 6:11 PM

Misdirection

dofmtoday at 6:54 PM

If Mythos is still running internally, the NSA still have some access to it. It's just crazy to believe there aren't CIA and/or NSA plants (tacitly acknowledged or otherwise) inside Anthropic and OpenAI.

But Mythos is still only an advanced LLM so I am not sure what all this breathy fuss is about; it sounds like the PR war more than anything.

If the NSA aren't themselves training technologies that are at least as powerful, that would modestly surprise me.

Not that you need an LLM to monitor the risks to the USA. You just need Tulsi Gabbard's emails.

show 1 reply
ransom1538today at 7:33 PM

NSA has produced nothing. Does nothing. Why don't we just have them pick up garbage on the freeway to help out the tax payers? Let Anthropic and other adults push spying forward.

show 4 replies
swader999today at 6:26 PM

Kind of crazy actually. Other models are catching up fast, they all can find the vulnerabilities in our (and by our I mean everyone's) underlying infra very fast. It takes a very long time to fix, review, and finally deploy these fixes. There really isn't much time left.