Anthropic recently took two of its latest AI models offline due to an export control order from the Trump administration, sparking a broader debate about AI policy and digital sovereignty.
On the latest episode of TechCrunch’s Equity podcast, Sean O’Kane, Rebecca Bellan, and I discuss what actually prompted the administration’s move against Anthropic and what this means for the broader AI ecosystem.
Anthropic’s other competitors may not have to worry about a similar crackdown, since, as Sean notes, “Anthropic hasn’t had the best relationship with the Trump administration in terms of what sets it apart from other major AI labs.”
However, Rebecca also noted that leading cybersecurity experts have “signed an open letter to President Trump asking him to rescind the order, saying it is actually dangerous to have to extract these advanced cybersecurity capabilities from America’s network defenders.”
And we thought this might be good publicity for Anthropic, especially since, in Rebecca’s words, “everyone loves a bad girl.”
Continue reading for a preview of the conversation, edited for length and clarity.
Rebecca Beran: As many of our listeners know, the U.S. government basically just forced Anthropic to take their two newest models offline, the Fable 5, and then they had the Mythos 5 as well. This is available to current Mythos users, (whereas) Fable 5 was available to the public.
They sent a letter (last week) Friday citing “national security concerns.” No one knows what those concerns are. That report has not been made public, and they did not give details, telling (Anthropic) that they needed to ensure those models were not used by foreigners. So Anthropic said, “Okay, I think we need to completely take over the model, because we don’t know if someone is foreign. A lot of our employees are foreign.”
But in fact (according to reports) the White House tipped off this because some Amazonian researchers had allegedly discovered a way to bypass Fable 5’s guardrails. Amazon CEO Andy Jassy raised these concerns with the White House, and things only got worse from there.
Sean O’Kane: Everything went really quickly, especially from Friday afternoon to the weekend. And that was at the same time that the administration was ostensibly trying to negotiate some kind of treaty regarding the war it had started in Iran.
Rebecca: It’s Friday evening for us in New York. They love distractions.
Sean: Let’s step back quite a bit for a moment. Anthropic doesn’t have as good of a relationship with the Trump administration as other big AI labs. At least, I think there are elements of that here.
So do you think this will affect other companies? Do you think the Trump administration won’t turn off the spigot on any of these competitors?
Anthony Ha: Part of the context here is that both reports and analysis by independent security experts suggest that the actual security risk posed by Anthropic is not that unique. So it seems like a lot of this problem stems from the Trump administration and some of Anthropic’s not getting along very well. Whatever risks there are, those things are going to blow up out of proportion just because they don’t seem to be able to make commercial phone calls to each other.
On the one hand, if you are in another company, it may be advantageous to you. Because we can say, “Well, we just don’t get mad at them and we can do what we want.” But this is also not a great regulatory situation to just say, “I hope they don’t get mad at us.”
Rebecca: On the one hand, this definitely feels retaliatory. A major lawsuit is brewing between the two countries after the government labeled Anthropic a supply chain risk. It feels like the White House is just looking for an excuse to attack Anthropic. I felt that way not only because that was my first reaction, but also because of what many cybersecurity researchers said. It says that export restrictions (orders) should not have been triggered as a result of this. They have all signed an open letter asking President Trump to rescind the order, arguing that it is actually dangerous to have to extract these advanced cybersecurity capabilities from US network defenders. Anthropic itself has also said that some of the same jailbreaks may have been found in several other AI models.
Ironically, it’s something like this: “Okay, are you just pausing Anthropic to let other people catch up on what’s going on with Anthropic?”
But at the same time, I’ve also seen reactions like this: (For example): Anthropic expected something like this to happen. They say, “This is too dangerous for anyone to use, but we’re not. We’re the good guys.” They speak out of both sides of their mouths. A week before Fable came out, they said, “We need to slow down the AI. It’s getting really dangerous.” But then it boomed, “This is our craziest, super-powerful model ever, go away.”
Anthony: In some ways, this feels like a microcosm of a lot of the debate around AI. There, people like Sam Altman and Jensen Huang are saying, “Let’s turn the temperature down. Why is everyone mad at us?” Well, you’ve spent the last few years essentially saying that you’ve built a God machine that will take jobs away from everyone. It’s not really shocking that people aren’t happy about this.
And when Anthropic specifically talks about Mythos, it’s like, “This is the most incredibly powerful model ever, and it’s too dangerous to release to the public.” And on some level, (you say), “Well, let’s say we take it seriously. That means there’s going to be an incredible level of scrutiny about it.”
And I’m wondering, it looks like Anthropic isn’t happy about this. I want to be careful not to overstate how beneficial this is for them. But we also published several articles on ramp analysis to highlight the fact that the last big clash between Anthropic and the Trump administration was, at least in some ways, good for the company. Claude’s download numbers skyrocketed. I think a lot of people who previously thought of ChatGPT as a chatbot, an AI assistant, suddenly started thinking of Claude as perhaps a more responsible, more “resilient” person.
And similarly, Anthropic is very stressed about this, but this could also make their model look even stronger.
Rebecca: Certainly. “We’re so dangerous.” Everyone loves a bad boy, right? Everyone is saying, “This is the most powerful model. Trump says so. Of course I have to have it.”
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
