AI
Trump administration's Anthropic crackdown sparks debate over actual AI risk
A forced pulldown of Anthropic's latest models raises questions about whether regulatory action reflects genuine security concerns or escalating tensions between the company and the administration.
TechCrunch — AI
The Trump administration recently ordered Anthropic to take its two newest AI models—Fable 5 and Mythos 5—offline citing undefined national security concerns, reigniting debate about whether the move addresses real AI risks or reflects political friction.
On a recent episode of TechCrunch's Equity podcast, Sean O'Kane, Rebecca Bellan, and Anthony Ha unpacked the incident. The trigger reportedly came after Amazon researchers discovered potential guardrail bypasses in Fable 5, with CEO Andy Jassy flagging concerns to the White House. The administration then issued a vague order requiring Anthropic to ensure foreign nationals couldn't access the models—forcing a complete pulldown rather than a targeted restriction.
A pattern of friction?#
What's notable is Anthropic's already-strained relationship with the administration. As O'Kane noted, tensions have been "apart from the other leading AI labs"—leaving competitors wondering whether they need to worry about similar enforcement.
Yet cybersecurity experts signed an open letter urging Trump to revoke the order, arguing that pulling advanced defensive AI capabilities from U.S. network defenders creates actual security gaps. Anthropic countered that similar jailbreaks could affect other models, suggesting the crackdown targets one company rather than a category of risk.
The paradox of being "dangerous"#
There's an inherent contradiction in Anthropic's positioning. The company spent months emphasizing Mythos's unprecedented power and danger, even as it released the model. That messaging—combined with a prior plea to "slow down AI"—may have invited regulatory scrutiny. As one panelist put it: if you say a model is too dangerous to release, you've given officials every reason to take that claim seriously and act on it.
Meanwhile, past crackdowns on Anthropic appear to have boosted its brand. Downloads of Claude spiked after previous tensions, as some users switched from ChatGPT seeking the "more responsible" alternative. This cycle could repeat—the forbidden fruit effect amplifying demand for models framed as powerful enough to warrant government intervention.
The broader question: does this landscape benefit Anthropic's rivals by lowering regulatory heat on them—or does it create an unstable environment where any AI company risks becoming a political target?
