Technology & Science
U.S. Export Controls Force Anthropic to Pull Fable 5 & Mythos 5 AI Models Worldwide
On 12 June 2026, after a 90-minute ultimatum, the Commerce Department invoked national-security export controls that compelled Anthropic to disable its just-launched Fable 5 and Mythos 5 models for any foreign user.
Focusing Facts
- The written directive arrived at 5:21 p.m. ET on 12 June 2026; Anthropic halted both models for all customers by 5:30 p.m. to comply.
- Amazon researchers had demonstrated that tailored prompts could make Fable 5 expose at least four previously known software vulnerabilities, triggering CEO Andy Jassy’s direct warning to senior White House officials.
- The ban coincides with Anthropic’s confidential IPO filing and extends even to the company’s own foreign-national employees, highlighting the breadth of the order.
Context
Washington has not slapped software itself with export controls this aggressively since 1997, when strong encryption was briefly treated as a munition under the Wassenaar regime. Today’s action signals a pivot from policing chips (e.g., the 2022–24 Nvidia/Huawei GPU restrictions) to policing the models that run on them—an echo of 1946’s McMahon Act, which shifted U.S. focus from uranium to the intellectual blueprints of the atomic bomb. The episode also exposes a perennial pattern: private industrial giants (here Amazon) using privileged access to steer security policy, much like AT&T’s role in 1934 radio regulation. On a century horizon, this could mark the moment when sovereigns began treating frontier AI weights as strategic materiel, foreshadowing a splintered global market where capabilities flow only within trusted blocs—potentially slowing the open-source diffusion that has characterized computing since the 1980s, but also birthing a new layer of techno-mercantilism that future historians may compare to early nuclear export regimes.
Perspectives
US national security–focused outlets
e.g., DT News, Asianet News Network, ANI — Frame the export-control order as a prudent move to block a serious “jailbreak” that could turn Anthropic’s Fable 5 into a cyber-weapon, stressing that national security must trump commercial goals. Echo the Trump administration’s America-First rhetoric and quote Pentagon officials approvingly, so readers mainly hear the government’s narrative while Anthropic’s rebuttals are downplayed.
Tech and crypto industry press
e.g., BeInCrypto, The Times of India business desk — Portray the shutdown as a heavy-handed, sudden government overreach triggered by Amazon, noting Anthropic’s claim that the flaws were minor and that such a standard would halt AI innovation. Lean toward protecting the start-up ecosystem and investors—emphasising IPO timing and innovation risks—so the security concerns may appear exaggerated or commercially motivated.
Political insider / Beltway publications
e.g., Politico report carried by Yahoo and Hindustan Times — Recount a frantic 24-hour tug-of-war between senior White House officials and Anthropic, highlighting Amazon CEO Andy Jassy’s call and tense phone diplomacy before export controls were imposed. Focus on intrigue within the corridors of power, which can sensationalise the drama and personalise policy disputes, potentially obscuring the technical details of the alleged vulnerability.
Like what you're reading?