US asks Anthropic to block global access to top AI models: Why it matters
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic disclosed on Friday that the US government issued an export control directive suspending all foreign national access to its Fable 5 and Mythos 5 models, rolled out less than a week earlier, including foreign national Anthropic employees and H1-B visa workers.
- David Sacks, a Trump adviser, said on X that the order stemmed from a warning that Fable 5 could be jailbroken, claiming CEO Dario Amodei was notified but "did nothing to fix it" — a characterization Anthropic publicly disputes.
- Anthropic said the government provided only "verbal evidence of a potential narrow, non-universal jailbreak" and called the situation a "misunderstanding," noting the company is working to restore access and that the directive must be applied "abruptly" across all customers.
- The directive adds a new front to Anthropic's existing lawsuit against the Trump administration, which placed the company on a supply chain blacklist after it refused to allow the US military to use its AI for domestic surveillance and fully autonomous weapons.
- Per a Semafor report dated June 14, the order was partly driven by suspicions that a China-linked group had accessed the new model, reflecting escalating US-China AI competition and Beijing's monopolization of rare earth supply chains critical to the sector.
- Experts told Reuters that Mythos models, in the wrong hands, could dramatically accelerate sophisticated cyberattacks — particularly in banking sectors that rely on decades-old interconnected systems.
- The block will disrupt firms like S&P, which integrates Anthropic's Claude software into databases used by financial advisers, investors, and analysts, while also affecting foreign researchers collaborating with the company.
Why it matters: Anthropic must abruptly disable Fable 5 and Mythos 5 for all customers to comply, hitting foreign nationals, H1-B workers, and enterprise clients like S&P that rely on Claude. The directive deepens Anthropic's existing lawsuit against the Trump administration over military AI use, turning a jailbreak dispute into a global access confrontation over frontier US models.
