On June 12, the US Commerce Department issued an export control order directing Anthropic to suspend access to its two most advanced AI models — Fable 5 and Mythos 5 — for all foreign nationals, including non-citizens working inside the United States and the company’s own employees. Anthropic disabled both models for every customer worldwide within hours.
Three weeks later, the motivations behind the order remain disputed, the legal basis unclear, and the question of who benefits from the crackdown far from settled.
The Commerce Department’s letter, signed by Secretary Howard Lutnick, cited “national security concerns” but did not make its full reasoning public. Multiple sources told Fox Business that the trigger was a jailbreak demonstration by Amazon AI researchers, who allegedly found a way to bypass Fable 5’s safety guardrails. Amazon CEO Andy Jassy raised the findings directly with the White House.
Anthropic disputed the significance of the jailbreak. The company stated that the vulnerability “is already available in other models including OpenAI’s GPT-5.5” — a model that remains fully operational and unrestricted. The implication was clear: if the criterion for an export control shutdown was the existence of a jailbreak, the government was applying it selectively.
The order forced Anthropic to shut down models that had been publicly available for only three days. The company said it could not guarantee that no foreign nationals would access them, given that many of its own employees are non-citizens working in the US.
A strained relationship
The crackdown did not emerge from a vacuum. Anthropic has had a notably confrontational relationship with the Trump administration that distinguishes it from other leading AI labs.
The administration had previously tried to prevent Anthropic from releasing Fable 5 at all. When that failed, it acted on the first plausible national-security justification it could find. A senior administration official told Fox Business that Anthropic’s “recklessness” in responding to safety concerns had eroded trust, and that the company’s CEO, Dario Amodei, was reportedly at a wellness retreat and unreachable during critical calls about the vulnerability.
Rebecca Bellan, reporting for TechCrunch’s Equity podcast, characterised the move as retaliatory: “After the government labeled Anthropic a supply chain risk, there’s this big lawsuit going on between them. It really feels like the White House is just looking out for any excuse to pummel Anthropic.”
Who benefits
The most immediate beneficiaries are Anthropic’s competitors. OpenAI’s GPT-5.5, which Anthropic pointed out had the same jailbreak vulnerability, remained available throughout. Meta’s Llama models face no similar restriction. Amazon, whose researchers triggered the crackdown, is both an investor in Anthropic and a direct competitor through its own AI initiatives.
The cynical interpretation, as Bellan put it: “Are you just pausing Anthropic so that others can catch up to where Anthropic was?”
Short-term competitive advantage is not the only benefit in play. The export control order arrived as the Trump administration was simultaneously negotiating over military actions in Iran. The timing allowed the White House to project strength on technology security during a period of broader geopolitical tension.
Foreign governments, however, are drawing different conclusions. The order has accelerated interest in non-US AI alternatives. French President Emmanuel Macron publicly signalled that France and India are exploring cooperative AI development, saying that the US tightening access to frontier models makes independent capabilities an urgent priority.
The unintended consequences
Seventy-six cybersecurity experts signed an open letter asking the president to revoke the order, arguing that pulling advanced AI capabilities from US network defenders makes the country less safe, not more.
The paradox is that if national security was the genuine concern, stripping the most advanced cybersecurity AI from the organizations that defend US infrastructure is a counterproductive way to address it. The experts’ letter noted that similar jailbreak techniques exist across multiple models and that shutting down one company’s offering does not eliminate the underlying vulnerability — it simply removes a defensive tool from the hands of those who need it.
The brand paradox
Paradoxically, the crackdown may benefit Anthropic in the long run. Ramp analysis cited by TechCrunch showed that the previous conflict between Anthropic and the Trump administration was good for business: Claude downloads surged as the company’s image as the “resistant” or “responsible” AI lab spread.
As Anthony Ha wrote: “Everyone loves a bad boy.” A government telling the world that an AI model is dangerous enough to ban is, for some customers, a powerful endorsement. And Anthropic’s own pre-launch messaging — warning that its models were too powerful for unrestricted release — created the narrative that the government has now ratified.
A dangerous precedent
The deeper concern is what the crackdown means for the regulatory landscape. If export controls can be triggered by an unverified jailbreak claim from a competitor, applied without a public evidence base, and used to shut down a product that a major company spent months developing, then every AI lab faces the same vulnerability. The rules are unclear, the process opaque, and the enforcement uneven.
Neither the Commerce Department nor the White House has made the full intelligence assessment behind the order public. Without transparency, the distinction between national security and competitive retaliation rests entirely on trust — and the events of the past three weeks have given the industry little reason to extend it.
Sources: TechCrunch (June 21); Fox Business (June 14); Creati.ai (June 16); TechCrunch cybersecurity open letter (June 15)