alt Hacker NewsU.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations
Comments
The real reason, afaik, that the US is trying to restrict access to SOTA models is that a very large component of USA tailored access and surveillance relies on exploits and weaknesses that these models will easily detect.
Thus, it really is an export control issue, but it has nothing to do with offensive capabilities. Offensive capabilities always exist, but pervasive defensibility would upset the asymmetric advantage that attackers, especially the USA, currently have.
There are now Asian models coming , optimized focused on cybersecurity defense at a high level, so I suspect this will be a relatively moot point soon.
LLMs are not great at creating exploits, but they are really good at detecting them. That asymmetry alone is enough to destroy the “offensive capabilities” narrative.
Yes, mythos can find exploitable bugs, even write bench exploits. But real exploits require a good dose of human psychology, and most of the tools needed are off the shelf available anyway. You still need a real cybersecurity expert to effectively weaponize a zero day into a deployable exploit.
But an LLM can inspect payloads, packages, and blobs en masse and find those exploits in a way that was wholly impractical before, so the asymmetric attack advantage is dissolved by strong LLMs.
The USA is trying to protect its cyberwarfare advantage, not protect against attackers. The exact opposite, actually. Porous security is a huge advantage to technologically advanced state actors.
> “I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model,” Commerce Secretary Howard Lutnick wrote to Anthropic’s chief compute officer Tom Brown Friday
why is the commerce secretary making this decision
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
>> More than 100 companies and institutions will now have access to Mythos 5, including many Fortune 500 companies, a source familiar with the new directive said, declining to be identified due to the sensitivity of the matter.
Who are those 100 companies? Clearly they can't compete on merit and have rubbed some hands to be picked as winners...at least for now.
I’m on record saying mistral needs 50B euros asap; I’ll admit this was wrong, they need more sooner
What’ll happen when Google Deep Mind go to release their next models, developed mostly in London?
Are Google going to end up in a situation where the people working on their models cannot use the models after launch?
This makes me sad since it implies that the best LLM I will ever be allowed to use is GPT 5.5 and Opus 4.8. Anything smarter than that is deemed too risky.
So much wasted potential.
And why would I pay Anthropic or OpenAI once consumer hardware gets powerful enough to run an open weight Chinese version of Opus 4.8? Even more so when mobile phones are able to run similar LLMs.
Their financial growth looks doomed. It looks like they will be heavily regulated just like the next missile factory. This is antagonist to VC led turbo growth startup regime.
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.
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.
Well, I think right now Antropic is going crazy with their safety rules. I got banned after creating a new account for a new employer. Went through some code with Claude Code, and that's... it. Came back the following morning and I was banned.
10 days for an appeal is too long. Company is a startup, so no team/enterprise support.
So what happens to their $1T valuations now, given they can't sell the product to consumers, and open-weight models are closing in on competition?
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 don't understand how this doesn't entirely screw the TAM for these companies?
As an EU company I think I now basically have to consider US AI as hostile and avoid it.
Dear anthropic, please consider moving your entire business to Canada or Europe where you will be allowed to conduct your business without vindictive and Kafkaesque government interference. I would like to have access to your models too!
Is there a world where frontier labs move to somewhere else like London to escape the business hurdles of the US? There are trillions at stake, is it a plausible scenario?
usgov picking winners and losers in AI --> usgov picking winners and losers in every industry (speedrun, any %)
This appears to be only for Mythos 5 access, NOT Fable 5.
This will be more validation that the rest of the world needs to be very wary of US technology.
It never made same sense that the most capable model was used by the CIA to create vault 7-like exploits while the same model was being used by another government project / random little people to patch up the vulnerabilities the exploits relied on.
So the general public get to bail-out & subsidize Anthropic / and downstream xAI/SpaceX via the State Wallet, but of course with no access.
Trusted US organisation: this must be the latest oxymoron.
I'm looking forward to better open source models. Now I just need to afford the compute to run these models.
This seems like it will have pretty huge negative affects on startups needing to compete with 'trusted partners'
The current digital hellscape that's developing is something we've been warned about consistently over the past 30 years.
All these dangers were known and predicted.
There's an uncanny parallel with the climate crisis.
Fatalistic somnambulism.
China will just buy a "trusted partner" one way or another.
It's like the epidemic of scam nvidia cards being resold without gpu or memory - where do you think those are going?
No one seeing this as Musk and SpaceX leveraging government support to close the IPO doors on OpenAI and Anthropic?
I know it’s a bold statement but look at this timing and their valuations going south.
"I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model"
I assume "trusted partners" means, "companies that have bribed Trump an appropriate amount". A few million for the inauguration, a few million for the ballroom, a few million on a movie about Melania, the don wants a taste.
It's a win-win game because both Anthropic and the Government are on the front page again pulling on important leavers.
In the mass-marketing world it's less about who's right or wrong but who is perceived by the population to be pulling the leavers on the front page again.
And we get the news the same time OpenAI releases 5.6. What a coincidence?
Hopefully this points toward Fable becoming available again.
[delayed]
It is interesting that there is no public announcement from the US government or Anthropic on this topic. That means there is no form to apply to be a trusted partner.
Does it mean US is allowing accessing to governments' exclusive list?
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).
Imposing a licensing system on models for limiting domestic use should require an act of congress but I mean obviously we're well past that red line.
Hard to see how this does not turn into export controls for models, just with a lot more ambiguity
Is there any scenario where it's not catastrophic for for the frontier labs?
They just got their market cut to a fraction. Investing in new tech is now very risky because even if things work out you might not be able to sell anything.
There were already serious doubts about ROI for the frontier labs. If they can only sell to 100 or so entities it's over business wise.
What's the endgame here?
Is there a list of the partners that get access? That should be public, right?
They play the game of China, they are stupid. US, as EU, lost everything with this stupid posture.
How would export controls apply if OpenAI or Anthropic released a model as open weights? Not that they would, but asking out of curiosity.
The funny thing is, they are planning to just use it for US government and not allow public to use fable 5 atleast..
Begun the AI wars have.
Next time someone tells you this is the party of free market and small government, I guess you just laugh now?
I can't wait till employees start helping with the distillation process.
ok, who in the EU is working on our own frontier model? surely we have the drive and ability?
It feels the U.S. is moving closer to a textbook definition of crony capitalism. Really sad but unsurprising with the current administration.
If this becomes the norm, what incentive does the rest of the world have to keep their markets open to the US?
If US companies have a large unfair advantage such that domestic competitors are no longer able to compete, then wouldn't it make sense for governments around the world to ban or tariff US products and services?
If I was responsible for national economic policy, I would place this at the top of my non-emergency agenda. The world needs to act quick before their industries fail.