The Trump administration recently imposed export controls on Anthropic's newly released AI models, Fable 5 and Mythos 5, after a frantic 24-hour effort to convince the company to voluntarily pull them over security risks [finstreetnews.com]. This decision followed multiple tense calls between Anthropic CEO Dario Amodei and administration officials, including Treasury Secretary Scott Bessent and White House Cyber Director Sean Cairncross, highlighting the White House's struggle to regulate rapidly evolving AI models [finstreetnews.com]. The export control order, received by Anthropic on Friday at 5:21 PM ET, banned the models' use by foreign nationals, but the company determined the only way to comply was to completely disable them for all customers worldwide [theverge.com, techcrunch.com].
Anthropic had released Fable 5 as a "Mythos-class model" with touted safeguards, believing it to be safe for general use, even after extensive reviews by the administration and the UK's AI Security Institute [finstreetnews.com, cautiousoptimism.news]. However, soon after its public release, top administration officials developed fresh doubts about the security of the AI's guardrails [thenyledger.com]. Anthropic stated that the government's concern stemmed from a "potential narrow, non-universal" method of bypassing, or "jailbreaking," Fable 5, which the company claims was shared with the government by an unnamed entity [theverge.com, techcrunch.com]. Anthropic further argued that the identified jailbreak was not unique to Fable 5 and that a similar "level of capability" is widely available in other models, including OpenAI's GPT-5.5 [theverge.com, techcrunch.com].
Despite Anthropic's reassurances and past efforts to position itself as a vocal advocate for AI regulation, the administration acted due to national security authority [thenyledger.com, finstreetnews.com]. Former White House AI czar David Sacks, while generally anti-regulation, supported the administration's decision, expressing bewilderment that Anthropic hadn't complied with safety requests it previously prioritized [finstreetnews.com, thenyledger.com]. He expressed hope that Anthropic would remediate the safety issue so the export control could be lifted and Fable 5 could return to general release [finstreetnews.com, thenyledger.com]. Anthropic's Mythos model had previously been restricted due to its ability to find security vulnerabilities, leading to a controlled program called Project Glasswing before Fable 5's release [techcrunch.com, cautiousoptimism.news].
