I've read up a little more on this. Anthropic's reach exceeded their grasp. They're attempting to deploy a potentially highly destructive technology with inadequate security measures. Gun dealers need an FFL license and to run background checks through the FBI. Chemical manufacturers have restrictions on what precursors are sold to who and in what quantities on top of chain of custody traceability. When it comes to an AI model that can autonomously hack very important infrastructure quickly and at a mass scale, there was no effing way that was going to get anything short of the highest degree of scrutiny. Now they need to develop safety measures for it that satisfy the government's concerns about this being used against the US by foreign actors.

ntxVol said:

---

Spergin said:

---

This is massive change but inevitable. ITAR laws are coming to AI now, it's pretty guaranteed.

---

That's pretty ridiculous, this would destroy AI research in the US.

---

Destroy AI Research or Destroy the country and the world?

We got a choice.

I am all for AI but with guard rails. If you can't build a nuke, you can't build dangerous AI either.

ntxVol said:

---

Spergin said:

---

This is massive change but inevitable. ITAR laws are coming to AI now, it's pretty guaranteed.

---

That's pretty ridiculous, this would destroy AI research in the US.

---

Why? China doesn't use anyone but Chinese people. We don't need foreigners.

Turns out it was Andy Jassy himself (Amazon CEO) who triggered this.

Quote:

---

The Trump administration's decision to halt all foreign use of Anthropic's most-capable AI models was prompted by conversations between Amazon.com Chief Executive Andy Jassy and U.S. officials including Treasury Secretary Scott Bessent, people familiar with the matter said.

The administration had long felt that Anthropic, one of the leaders in America's artificial-intelligence race, couldn't be trusted to manage the security risks its new model presented. Friday's call between some administration officials and Anthropic Chief Executive Dario Amodei reinforced that idea, the people said. Some of the officials said the company indicated that it was unwilling to work with security experts in the government to address the issue.

In response to the government's move Friday, Anthropic said it was shutting off access to Mythos and Fable to all users to make sure it complied, potentially hampering efforts by companies around the world to use the tools to identify software vulnerabilities. Many of the researchers at Anthropic are foreign-born, meaning the government's rule effectively prevented them from working on the latest models, the company said.

---

Based on this article and another I read yesterday it feels like Anthropic is the issue here.

Quote:

---

The senior White House official said the new intervention was about model safety and doesn't detract from the administration's innovation focus. White House AI adviser and venture capitalist David Sacks, an Anthropic critic, said on social media that the administration had issued the restriction "reluctantly."

The administration's "hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release," he wrote.

---

Spergin said:

---

ntxVol said:

---

Spergin said:

---

This is massive change but inevitable. ITAR laws are coming to AI now, it's pretty guaranteed.

---

That's pretty ridiculous, this would destroy AI research in the US.

---

Why? China doesn't use anyone but Chinese people. We don't need foreigners.

---

but but but free markets.... capitalism.... conservatives.... need overseas "talent".. not xenophobic...

heh.

As days go by, people are realizing that these are all just empty buzzwords that rich oligarchs used to tether common folk to lifelong slavery and labor while they got richer themselves.

YouBet said:

---

Turns out it was Andy Jassy himself (Amazon CEO) who triggered this.

Quote:

---

The Trump administration's decision to halt all foreign use of Anthropic's most-capable AI models was prompted by conversations between Amazon.com Chief Executive Andy Jassy and U.S. officials including Treasury Secretary Scott Bessent, people familiar with the matter said.

The administration had long felt that Anthropic, one of the leaders in America's artificial-intelligence race, couldn't be trusted to manage the security risks its new model presented. Friday's call between some administration officials and Anthropic Chief Executive Dario Amodei reinforced that idea, the people said. Some of the officials said the company indicated that it was unwilling to work with security experts in the government to address the issue.

In response to the government's move Friday, Anthropic said it was shutting off access to Mythos and Fable to all users to make sure it complied, potentially hampering efforts by companies around the world to use the tools to identify software vulnerabilities. Many of the researchers at Anthropic are foreign-born, meaning the government's rule effectively prevented them from working on the latest models, the company said.

---

Based on this article and another I read yesterday it feels like Anthropic is the issue here.

Quote:

---

The senior White House official said the new intervention was about model safety and doesn't detract from the administration's innovation focus. White House AI adviser and venture capitalist David Sacks, an Anthropic critic, said on social media that the administration had issued the restriction "reluctantly."

The administration's "hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release," he wrote.

---

---

So Jassy got gassy and ruined it for everyone else?
ha ha ha.

Ultimately the Government is the boss. Smart leaders work with them, not take them on. Amodei is in trouble for messing with Trump and acting high and mighty to impress his woke friends.

Send all those foreign people away, they can help their own countries with their "skills" and "talent".

.... and extend the same rule to ALL US companies.

Defense companies can't sell to any country they desire. American AI should not be sold to anybody.

Dan Scott said:

---

Defense companies can't sell to any country they desire. American AI should not be sold to anybody.

---

Agreed. Like I said earlier this is self-fulfilling prophecy.

If we are to believe their own words and predictions, then we simply can't let these companies release this stuff to the wild or at least to foreigners. This mindset that AI is going to take over and kill us all, and someone is going to release it so it might as well be us, is cracked thinking.

How do they expect the government to respond to this? I'm gonna actually defend the government for once on this issue.

There is more to the story here. Either the vulnerabilities weren't serious (which matches Anthropic's "previously known, minor" line and undercuts the emergency), or they were serious and the Glasswing partners - Amazon foremost - failed at the exact defensive job the program existed to do.

infinity ag said:

---

ntxVol said:

---

Spergin said:

---

This is massive change but inevitable. ITAR laws are coming to AI now, it's pretty guaranteed.

---

That's pretty ridiculous, this would destroy AI research in the US.

---

Destroy AI Research or Destroy the country and the world?

We got a choice.

I am all for AI but with guard rails. If you can't build a nuke, you can't build dangerous AI either.

---

So the government has made the choice, the US will not be a competitor in AI. Got it.

I believe if the government could force Anthropic to block access completely they would have done so. An ITAR classification is the most they can legally do at this moment.

Effectively though, it has blocked it for everyone. It's not an available model right now, Anthropic doesn't have a way to verify US citizenship at this moment, so its offline for all general customers at this point.

Which sucks, I was planning to start using it on Monday at work. Lesson learned, when this stuff comes out use it immediately before it gets blocked.

I don't have a good answer on this. I don't like what just happened, but I don't see a good path going forward. These tools are so powerful, they make any body with a laptop and internet access capable of extremely widespread damage. On the other hand, they can't be stopped, it's not like the US is the only country working on this, and the speed at which these models are growing makes it likely that an LLM with Mythos capabilities is downloadable for anybody for free within the next 12 months

This is really spiraling quickly

ntxVol said:

---

infinity ag said:

---

ntxVol said:

---

Spergin said:

---

This is massive change but inevitable. ITAR laws are coming to AI now, it's pretty guaranteed.

---

That's pretty ridiculous, this would destroy AI research in the US.

---

Destroy AI Research or Destroy the country and the world?

We got a choice.

I am all for AI but with guard rails. If you can't build a nuke, you can't build dangerous AI either.

---

So the government has made the choice, the US will not be a competitor in AI. Got it.

---

Not being a competitor vs destroying the world, I will pick the former.

But in reality, it is not that bad. There is a middle ground. I know you may not like guardrails on corporations but there need to be rules for everyone.

If any of these corps build AI to destroy the world (if that is possible), and sell it to some terrorist, throw them into prison.

You know what else is dangerous? Good encryption. Let's also make that illegal. We don't want the commies to be able to hide stuff so make it impossible for anyone to use without registering their encryption keys.

You guys (and the DJT admin) sound like a bunch of olds asking their grandkids "do you know how to post videos to Facebook?"

kb2001 said:

---

I believe if the government could force Anthropic to block access completely they would have done so. An ITAR classification is the most they can legally do at this moment.

Effectively though, it has blocked it for everyone. It's not an available model right now, Anthropic doesn't have a way to verify US citizenship at this moment, so its offline for all general customers at this point.

Which sucks, I was planning to start using it on Monday at work. Lesson learned, when this stuff comes out use it immediately before it gets blocked.

I don't have a good answer on this. I don't like what just happened, but I don't see a good path going forward. These tools are so powerful, they make any body with a laptop and internet access capable of extremely widespread damage. On the other hand, they can't be stopped, it's not like the US is the only country working on this, and the speed at which these models are growing makes it likely that an LLM with Mythos capabilities is downloadable for anybody for free within the next 12 months

This is really spiraling quickly

---

I used it before they pulled it. I then found out that for Fable, Anthropic was storing data back to their servers - to be deleted after 30 days after review - but not using it to train. I immediately stopped using it at that point.

I tried to get Fable to audit my code for security vulnerabilities. About half way thru it dropped me back to Opus 4.8 with some generic warning. No idea.