• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

U.S. Loosens Restrictions on Anthropic’s Mythos A.I. Model

P

philb2

2[H]4U
Joined
May 26, 2021
Messages
3,471
https://www.nytimes.com/2026/06/26/technology/anthropic-mythos-government-restrictions.html

Anthropic and the Trump administration reached a deal on Friday to bring one of the company’s most powerful artificial intelligence models back online, after days of tense negotiations.

Talks were still underway to restore access to another powerful Anthropic model, Fable 5, and to provide Mythos to more organizations, said a person familiar with the talks, who spoke on the condition of anonymity because the person was not allowed to discuss them publicly.

Earlier on Friday, OpenAI, Anthropic’s main rival, unveiled an A.I. technology called GPT-5.6 Sol, saying it matched the performance of Mythos on standard benchmark tests used by A.I. companies. OpenAI said it would initially share the technology with only a small number of companies approved by the administration.
 
This is the short-term future of AI in the USA. If your model is "too good" you'll be held back until other people catch up and then they'll let you release it.
 
Ididar said:
This is the short-term future of AI in the USA. If your model is "too good" you'll be held back until other people catch up and then they'll let you release it.
Click to expand...
The real danger is China is distilling the latest US models to get 85-90% of the performance of the top US models. US companies are spending trillions of dollars for others to copy them for a fraction of that amount. It's simultaneously both damaging to the US economy and a cybersecurity threat. US government may be trying to manage this exposure by limiting their access.
 
sleepeeg3 said:
The real danger is China is distilling the latest US models to get 85-90% of the performance of the top US models. US companies are spending trillions of dollars for others to copy them for a fraction of that amount. It's simultaneously both damaging to the US economy and a cybersecurity threat. US government may be trying to manage this exposure by limiting their access.
Click to expand...
IP theft by China is nothing new.
 
sleepeeg3 said:
The real danger is China is distilling the latest US models to get 85-90% of the performance of the top US models. US companies are spending trillions of dollars for others to copy them for a fraction of that amount. It's simultaneously both damaging to the US economy and a cybersecurity threat. US government may be trying to manage this exposure by limiting their access.
Click to expand...

Possible, although given this administration's eye towards bribery and punishing people they don't like, I suspect it has very little to do with China. Every time they've talked China a well placed bribe has made those restrictions go away.

Even so, that only holds them back by months. That's hardly going to stop the issue from occurring. And, frankly, I don't really give a shit about people stealing AI because they're all based on stolen data anyway so I have no sympathy at all for the AI companies.
 
Ididar said:
I don't really give a shit about people stealing AI because they're all based on stolen data anyway so I have no sympathy at all for the AI companies.
Click to expand...
anthropic edge over China model is not built over stolen data, in coding there is giant open source available, open internet to read to train (anthropic robots do tend to respect robots.txt instruction) on and for Anthropic years of non api Claude-opus users data has well.

And for a long time now, synthetic data they get better and better at generating and of course a giant amount of model quality is not from data but reinforcement learning work, with paid humans or the best you can make AI feedback and when possible outcome based loop), China have access to all that stolen data, that not why they need to use frontier model in their training. A lot of it is for the reasoning part, not data.

Ididar said:
This is the short-term future of AI in the USA. If your model is "too good" you'll be held back until other people catch up and then they'll let you release it.
Click to expand...
It will continue to be more, held back until big player patched their code using that the "too good' model can found before anyone can use it and held back if someone like Amazon warn the authorities that the model seem to get broken to do dangerous stuff by people. I think a know your customer with approved list will probably be the future until other people catch up part.
 
LukeTbk said:
anthropic edge over China model is not built over stolen data, in coding there is giant open source available, open internet to read to train (anthropic robots do tend to respect robots.txt instruction) on and for Anthropic years of non api Claude-opus users data has well.

And for a long time now, synthetic data they get better and better at generating and of course a giant amount of model quality is not from data but reinforcement learning work, with paid humans or the best you can make AI feedback and when possible outcome based loop), China have access to all that stolen data, that not why they need to use frontier model in their training. A lot of it is for the reasoning part, not data.
Click to expand...

I get that. But the core was built on scraped data so as far as I'm concerned they can all bugger off when it comes to moaning about their work being stolen. I would have similar sympathy for a guy who robbed my TV having his computer stolen. None.
 
You must log in or register to reply here.
Back
Top