Live data from Hacker News

U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

semafor.com

311–320 of 873 posts

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#311

I’m not sure what the US government is trying to do. At first it seems like they are just trying to stifle some company that said no. Now they are just doing free publicity. It’s like never before have I wanted to try something out as much as this. They’re in effect saying “nothing else is as powerful as what Anthropic put out”. Even though that might not really be the case it’s what it sounds like.

Is there any evidence at all that would convince you that they're trying to mitigate real risks that actually exist?

Yes.

A published policy with the right to appeal exclusions from the list.

An equal standard for all companies rather than ad hoc application.

A countervailing policy to mitigate the unfair advantage conferred on the companies that have early access (such as a higher tax rate that goes to fund ai job loses, and a commitment that AI use of the new models won’t result in layoffs).

A requirement that hardware is made available for open source models rather than locked up in by the AI labs.

A restriction on AI labs being vertically integrated from hardware all the way up through the app layer. I would restrict AI labs to being API providers and prohibit them from building apps. That would allow an ecosystem of independent software development on the app layer without fear of being copied by the labs that have an unfair advantage in seeing the data while apps are being built, the usage data as they become successful and the ability to undercut competitors by subsidizing tokens unfairly.

I could go on.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#312

Don't start to rely on it . The US might remove access next month in a fit of pique. The Chinese models look increasingly more reliable and safer.

Safer?

If I can run it locally, it is certainly more safe than running on some cloud server somewhere.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#313
post #154

Don't start to rely on it . The US might remove access next month in a fit of pique. The Chinese models look increasingly more reliable and safer.

This is a pithy internet comment, but terrible advice. Between the Chinese government and Anthropic, I know which one I'd rather send tokens to. For all of the problems of the US, for-profit corporations, data harvesting, etc. the CCP (and, perhaps more troublesome, its allies) is far less likely to align with your interests.

Except the open weights can live on your hardware so no one gets your tokens.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#314
post #59

I understand why Anthropic might not want to fight this particular one in court, because they're trying to convince the administration to let them move forward. But would another company who is not on the trusted partner list and has less to lose taking on the admin have standing to sue here? On the basis of the export control being illegal and this putting their business at a disadvantage vs. competitors with access

Technically the US government is allowing Anthropic to serve the models to any US citizen, and it's Anthropic who decided that's impossible to comply with and so they pulled the model for everyone. I guess a US business with non-citizen employees could work.

A lawsuit would be a hard sell though, because Anthropic themselves argued that the technology is dangerous. Even if many people on HN might think that Anthropic was scaremongering about Mythos, a court is probably going to take their assessment at face value, and courts are loathe to find against governments in cases of national security.

There's also the issue that these models are getting better through an iterative process, so even if the line between GPT 5.5 and Fable/GPT 5.6 is somewhat arbitrary, it doesn't mean that the government shouldn't be able to draw a line at all. So you're left arguing that they drew the line too early, which is subjective.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#315

I’m not sure what the US government is trying to do. At first it seems like they are just trying to stifle some company that said no. Now they are just doing free publicity. It’s like never before have I wanted to try something out as much as this. They’re in effect saying “nothing else is as powerful as what Anthropic put out”. Even though that might not really be the case it’s what it sounds like.

Grift. It's all about the grift.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#316
Are Cyber Verification Program (CVP) members included in this?: "We also intend to scale up our Cyber Verification Program, which would grant Mythos-class capabilities to many more organizations for specific cyberdefense tasks" (https://www.anthropic.com/news/expanding-project-glasswing).

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#317
post #101

I wonder if the Founding Fathers knew about AI, they would include it in the 2nd? The spirit is to provide effective tools for the people to resist federal military tyranny, and Mythos seems like it would be a good tool to defend against that, for so many reasons.

I'm not sure that analogy works: pretty much everyone agrees that there are some types of weapons civilians shouldn't be able to have, even though they might be very effective for resisting military tyranny.

I don’t think everyone agrees on that. It’s just that the government has been able to legislate that as technology evolved.

That being said many legal scholars say the state militia was intended to be the defense against tyranny not individual citizens because there were government led crackdowns on rebellion under Washington and other presidents from the earliest days of the republic. State militias have the full range of weapons

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#318
post #283

Earlier quoted context omitted.

Yes he does. They could ignore the US government, but will likely quickly find themselves in court fighting a fight that they are likely to lose and isn’t worth fighting anyway.

[flagged]

You are not wrong but I would also consider that it's relative easier to sentence someone who is nobody than someone famous in that regard.

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#319
post #59

I understand why Anthropic might not want to fight this particular one in court, because they're trying to convince the administration to let them move forward. But would another company who is not on the trusted partner list and has less to lose taking on the admin have standing to sue here? On the basis of the export control being illegal and this putting their business at a disadvantage vs. competitors with access

Technically the US government is allowing Anthropic to serve the models to any US citizen, and it's Anthropic who decided that's impossible to comply with and so they pulled the model for everyone. I guess a US business with non-citizen employees could work. A lawsuit would be a hard sell though, because Anthropic themselves argued that the technology is dangerous. Even if many people on HN might think that Anthropic…

Are you certain? Trump admin is hand-picking GPT 5.6 winners

Re: U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations

#320
post #61

Earlier quoted context omitted.

Other than maybe some in-the-moment cybersec wrappers, is this really true? Does anyone think a startup with a good product is going to be materially disadvantaged by not having access to an incrementally better security focused LLM release? It’s lots of fun to pretend it’s some step-change that’s too dangerous for general release, but in real life it’s not conferring some massive advantage that any real startup woul…

> Does anyone think a startup with a good product is going to be materially disadvantaged by not having access to an incrementally better security focused LLM release? - It's not "incrementally better". It's a complete game changer. Opus 4.8 on max thinking does X amount of mistakes in my commercial work. Fable 5 did 5% of X. Counted. I barely had anything to contribute in the work sessions, for a full week I could c…

> I don't have too much unknowns in my work, I simply have way too much to do.

Interesting, I'm curious what work you do? My software engineering career has never been in that situation, it's always so much ambiguity and unknown that trumps everything.

Post reply on HN