Why did the Trump administration force Anthropic to take down its latest AI models?

Anthropic has been compelled to take down its two latest AI models, Fable 5 and Mythos 5, following an export control directive issued by the Trump administration, which cited national security concerns without disclosing specific details. This move has ignited extensive debate about the future of AI regulation and digital sovereignty.

During a recent episode of TechCrunch’s Equity podcast, Sean O’Kane, Rebecca Bellan, and Anthony Ha examined the motivations behind the administration’s action against Anthropic and the potential consequences for the wider AI industry.

Rebecca Bellan explained that the U.S. government’s letter, delivered last Friday, mandated that Anthropic prevent any foreign nationals from accessing these models. Given the difficulty in verifying users’ nationalities and the presence of many foreign employees within Anthropic, the company opted to remove the models entirely from availability.

Reports suggest that concerns arose after Amazon researchers allegedly discovered a method to bypass Fable 5’s safety guardrails. Amazon CEO Andy Jassy reportedly alerted the White House, which escalated the situation rapidly over a weekend.

Sean O’Kane noted the swift progression of events, coinciding with the administration’s ongoing efforts to negotiate a treaty regarding the conflict it initiated in Iran. He also highlighted that Anthropic’s relationship with the Trump administration has been notably strained compared to other leading AI labs.

Anthony Ha remarked that independent security analyses indicate the risks posed by Anthropic’s models are not uniquely significant. He suggested that tensions between the administration and Anthropic may have amplified the response, potentially benefiting competing companies that maintain better relations with regulators.

Rebecca Bellan characterized the government's actions as potentially retaliatory, pointing to a lawsuit between Anthropic and the administration and the labeling of Anthropic as a supply chain risk. She referenced an open letter signed by cybersecurity experts urging the Trump administration to revoke the order, warning that removing these advanced cybersecurity tools from U.S. network defenders could be dangerous. Anthropic itself acknowledged that similar vulnerabilities exist in other AI models.

Bellan also raised the possibility that the export control could serve to delay Anthropic’s progress, allowing competitors to catch up. She noted the contradictory stance of Anthropic, which had publicly called for slowing AI development due to safety concerns shortly before releasing its most advanced model.

Anthony Ha compared this situation to broader AI debates, mentioning figures like Sam Altman and Jensen Huang who have called for reducing tensions around AI. He emphasized that the portrayal of AI as an overwhelmingly powerful and disruptive technology has contributed to public unease.

Ha pointed out that Anthropic’s own description of Mythos as an extremely powerful and potentially dangerous model invites heightened scrutiny. Despite the company’s apparent displeasure with the export control, previous conflicts with the Trump administration have paradoxically increased interest in Anthropic’s products, with downloads of its Claude model rising.

Rebecca Bellan concluded that the controversy might inadvertently enhance Anthropic’s reputation, as public fascination with the “dangerous” nature of its models grows. She remarked, “Everyone loves a bad boy,” suggesting that the perception of Anthropic’s models as powerful and restricted could increase demand.