Washington’s Anthropic Crackdown Sends Shockwaves Through Global AI
A single U.S. government order has dominated the AI agenda over the past day, forcing Anthropic to disable two of its most capable models worldwide and reigniting a global debate about who really controls frontier AI. Reliability concerns added a separate, awkward storyline of their own.
U.S. directive forces Anthropic to pull two frontier models
Anthropic suspended global access to its Fable 5 and Mythos 5 models after a Commerce Department order, signed by Secretary Howard Lutnick, demanded the company cut off all foreign-national access on national-security grounds. Rather than build a partial block, Anthropic disabled the models entirely for every customer, an unusually blunt response to a government mandate. The episode is among the most direct interventions yet by Washington into a leading lab’s product line. (CNN Business)
White House ties the curbs to China and resists widening them
Officials linked the restrictions to suspicions that a China-linked group had accessed Mythos, alongside a warning that Fable 5 could be jailbroken. According to reporting, the controls followed several tense calls between CEO Dario Amodei and administration officials. For now, a source says the White House is unlikely to extend similar export limits to other AI companies. (Semafor)
India confronts its dependence on foreign AI
The suspension landed hard in India, Anthropic’s second-largest market, where firms such as TCS and Infosys had built strategies around its models. The disruption has revived calls for “sovereign AI,” with Zoho founder Sridhar Vembu urging local organizations to lean on smaller, open-source systems they can control. (TechCrunch)
KPMG pulls an AI report riddled with AI hallucinations
In an unrelated but telling stumble, KPMG withdrew its flagship report “Redefining excellence in the age of agentic AI” after organizations including UBS and the UK’s NHS said claims about their AI use were fabricated. Researchers found that of 45 citations, only five pointed to genuine sources, with the errors traced to AI hallucinations. (TechCrunch)
What to watch: whether other governments follow Washington’s lead on frontier-model export controls, and how quickly the “sovereign AI” push translates into real investment outside the U.S.