Claude Mythos 5 cleared again – but only for selected US partners (Fable 5 stays blocked)
Two weeks after the sales ban, the US government is partly relenting: Anthropic may again provide its strongest AI model, Claude Mythos 5 – but only to a fixed group of around 100 US organizations, primarily operators and defenders of critical infrastructure. The permission does not cover the weaker sister model, Fable 5.
The basis is a letter from Commerce Secretary Howard Lutnick to Anthropic's chief compute officer Tom Brown on Friday, first reported by Semafor and Reuters. For the recipients listed in an annex to the letter, the previous export-license requirement is dropped – explicitly including their foreign employees as well as Anthropic's own non-American staff, who had also been locked out by the original ban.
The government had cut access to both models on 12 June, after security researchers apparently bypassed the built-in safeguards with ease. Fable 5 had been made publicly available a few days earlier as a supposedly better-protected variant – and is now the part that remains blocked.
Anthropic confirmed the step and said it would quickly restore access for the affected organizations and continue working with the government to widen the circle and make Fable 5 generally available again. When that will happen is open.
Almost simultaneously, OpenAI released its new GPT-5.6 – likewise only in limited form and after coordinating with the government (more in our article on GPT-5.6). We described the background to the original ban in our article „Anthropic, Mythos and the 3-day wonder Fable".
Sources
Handle invoices more easily
Easy Invoice combines quotes, invoices and customer management in the cloud.
Try Easy InvoiceRead next
AI & Tools
OpenAI Jalapeño: half the cost per AI request – will AI get cheaper or will prices rise anyway?
OpenAI promises roughly half the cost per token with its Jalapeño chip. Will that make AI cheaper – or will prices rise anyway? An honest assessment.
AI & Tools
GPT-5.6 is here – but only for a few: Sol, Terra and Luna in a limited launch
OpenAI has unveiled GPT-5.6 – but almost no one can use it. The three models Sol, Terra and Luna launch only for about 20 government-approved partners at first.
AI & Tools
Anthropic, „Mythos" and the 3-day wonder „Fable": when the US government switches off an AI
On 9 June 2026 Anthropic launched its most powerful public AI model, Fable 5 – the safety-hardened version of the held-back „Mythos". Three days later a US government export-control directive forced the company to disable Fable 5 and Mythos 5 worldwide. Clever marketing or something more? An assessment.