After weeks of back-and-forth with the Trump administration, Anthropic has announced that its long-sidelined Claude Fable 5 will be restored to users globally. The company says it is working on improved safeguards and plans to share more information with governments.

The return comes after early June’s export control directive from the Trump administration, which halted access due to concerns over potential jailbreaks of the technology. Anthropic has since trained a new safety classifier that blocks such behavior in over 99% of cases.

In a bid for government approval, Anthropic now plans to offer pre-release access and evaluation for models relevant to national security capabilities. They also intend to introduce rapid information sharing when significant jailbreaks or misuse patterns are identified. The company is collaborating with other leading AI labs to create a shared, voluntary security and evaluation standard.

Despite this progress, the initial export control directive came at an inopportune time for Anthropic, which was preparing for its IPO and had been feuding with the government over supply chain risk designations. The company highlights that there is currently no consensus on how to assess jailbreaks' severity within the AI industry.

Anthropic hopes to address this through partnerships with Amazon, Microsoft, Google, and others in Project Glasswing, drafting a framework for assessing jailbreaks based on their impact. They will also monitor key submission channels 24/7 and debut a HackerOne program to engage researchers.