Anthropic recently took two of its new AI models offline due to an export control order from the Trump administration, sparking broad debates about AI policy and digital sovereignty.
In a recent episode of TechCrunch’s Equity podcast, Sean O’Kane, Rebecca Bellan, and I discussed what exactly caused the management move against Anthropic, and what this might mean for the broader AI ecosystem.
As Sean put it, “Anthropic hasn’t had the best relationship with the Trump administration in a way that other top AI labs have,” so perhaps Anthropic’s other competitors don’t need to worry about the same crackdown.
But Rebecca also noted that leading cyber security experts “signed an open letter asking Trump to withdraw the order, and they say it’s really dangerous to release these advanced cybersecurity skills to US network defenders”
And we wondered if all this could end up being a good thing for Anthropic, especially since – in Rebecca’s words – “everyone loves a bad boy.”
Read on for a preview of our interview, edited for length and clarity.
Rebecca Bellan: As I’m sure many of our listeners know, the US government basically forced Anthropic to pull two of its new models offline – the Mythos 5, and then the Mythos 5, which was available to current Mythos users. [whereas] Legend 5 was widely available to the public.
They sent a letter [last] Friday citing “national security concerns.” No one knows what those problems are. That report has not been disclosed, they did not give specific details and were told [Anthropic] that they had to ensure that those species could not be used by any outsiders. So Anthropic was like, “Okay, I think we should just take the models out completely, because we don’t know if someone is foreign. Most of our workers are foreign.”
But actually, [reports said] The White House was alerted to this because some Amazon researchers allegedly found a way to bypass Fable 5’s guardrails. Amazon CEO Andy Jassy raised this concern with the White House, and it just took off from there.
Sean O’Kane: All of this goes by very quickly, especially on Friday afternoon and it was the weekend. And that’s when the administration was clearly trying to negotiate some kind of deal for the war that started in Iran.
Rebecca: Friday evening for us in New York. They like to be disturbed.
Sean: Let’s backtrack a bit. Anthropic hasn’t had the best relationship with the Trump administration in a way that other leading AI labs have — I think there’s something, at least, of that at play here.
So do you think this will affect those other companies? Do you think the Trump administration would be more inclined to turn off the faucet on one of its competitors?
Anthony Ha: Part of the context here is that both reporting and analysis from independent security experts suggest that the actual security risk from Anthropic is not that different. So a lot of this seems to be largely coming from parts of the Trump administration and just Anthropic [not getting] very well. Whatever dangers there are, those things will explode in proportion because it seems that they cannot have a public telephone.
If you’re another company – on the other hand, maybe that’s beneficial to you, because you can say, “Well, we don’t annoy these people and we can do what we want.” But that’s also not a good place to just control [say]”Boy, I hope they don’t get mad at us.”
Rebecca: On the other hand, it feels vindictive – after the government called Anthropic a supply chain risk, there is this big lawsuit going on between them, it feels like the White House is looking for any excuse to oppose Anthropic. And I feel that way not only because that was my initial reaction, but because of what many cybersecurity researchers say. They say this should never have resulted in export controls [order]. They have all signed an open letter asking Trump to withdraw the order, and they say that it is actually dangerous to pull these advanced cybersecurity skills from the US network defenders Anthropic himself said that some of the same prison incidents can be found in several other types of AI.
Ironically, it’s like: OK, are you stopping Anthropic so others can get to where Anthropic was?
But at the same time, I have seen that reaction [say]: Anthropic kinda had this coming. They’re like, “This is too dangerous for anyone to use, but not us, we’re good guys.” They talk out of both sides of their mouth. A week before Fable came out, they were there [saying]”Hey, we need to take the AI down, guys. It’s getting really dangerous.” But then you say, “Here’s our craziest, most powerful model, go.”
Anthony: In some ways this feels like a microcosm of the many conversations around AI, where people like Sam Altman and Jensen Huang [saying]”Hey, let’s try to lower the temperature. Why is everyone mad at us?” Well, you’ve spent the last few years saying that God has this machine that will take jobs away from everyone. It’s not really a shock that people don’t feel good about this.
And there’s something about the way Anthropic talks about the Mythos in particular, where it’s like, “This is the most powerful model ever seen, it’s too dangerous to be released to the public.” And so on another level, [you say,] “Well, well, let’s just say we take that seriously. That means there’s going to be an incredible level of scrutiny around you.”
And I’m surprised – it seems that Anthropic is not happy about this. I want to be careful not to overstate how this would benefit them. But we also ran some stories about Ramp analysis to highlight the fact that the last big blowout between Anthropic and the Trump administration was good for the company, at least in some ways. Claude’s downloads increased. I think most people who have ever thought of ChatGPT as i chatbot, i The AI assistant before, they suddenly looked at Claude as the most responsible, the most “resistant”.
And in the same way, [while] Anthropic is very stressed about this, this, too, can make their models seem more powerful.
Rebecca: Definitely. “We are very dangerous.” Everyone loves a bad boy, right? Everybody’s like, “It’s a very powerful model, even Trump says so. Yeah, I’ve got to get my hands on it.”
If you shop through links in our articles, we may earn a small commission. This does not affect our editorial independence.
