Anthropic has reinstated customer access to its advanced AI model, Fable, following a suspension of over two weeks. This suspension was brought on by U.S. government concerns regarding potential misuse that could lead to serious cyber-attacks. The San Francisco-based company indicated that export controls had been lifted, referencing a statement by U.S. Commerce Secretary Howard Lutnick, who mentioned, “Over the past two weeks, we have worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the U.S. Government and strengthen America’s leadership in AI.”
Access to Fable 5 was restricted just three days after its launch on June 9 due to fears that both Fable and a related model, Claude Mythos 5, could be exploited by foreign adversaries. The U.S. government mandated an immediate ban on their usage by foreign nationals, regardless of their location, to avert potential cyber threats.
On Friday, the government partially lifted the restrictions by allowing Anthropic to distribute Mythos 5 to select “trusted” U.S. organizations, focusing on its use for defensive cybersecurity purposes. In a letter to Anthropic, which was reviewed by Reuters, Lutnick stated that the company had agreed to proactively address security risks associated with its models, collaborate with the government on standards for future releases, and report any malicious activities.
This ban reflects a shift in the White House’s approach to regulating AI companies, moving away from a previous “light touch” strategy aimed at maintaining competitiveness against China in the AI landscape. The government’s decision followed reports that Amazon researchers had discovered vulnerabilities in Fable 5 that could be exploited, raising alarms about its potential use in cyber-attacks.
In response to the export controls, Anthropic initially halted all access to the models but has since implemented enhanced safeguards. The company noted in a blog post that the safeguards have been tested and validated by the U.S. Department of Commerce’s Center for AI Standards and Innovation (CAISI), which affirmed their robustness.
Additionally, Anthropic announced its intentions to increase collaboration with government entities and establish dedicated teams to work on shared priorities. Notably, intelligence agencies from Australia, the U.S., the U.K., New Zealand, and Canada recently issued a warning about AI’s dual nature in cybersecurity, indicating that while it could bolster defenses, it also heightens the risk of sophisticated cyber threats.
The Five Eyes agencies noted that advancements in frontier AI models are expected to transform both offensive and defensive cyber capabilities considerably, predicting significant changes within months rather than years. Susie Wiles, chief of staff to former President Trump, expressed gratitude to companies engaging with the White House to implement executive orders promoting AI innovation and security, emphasizing a collaborative approach between the government and private sector.
However, the vetting process for which companies may access these powerful AI models has faced criticism. Sam Altman, CEO of OpenAI, voiced his concerns, stating that while safety testing is important, the notion of the government selecting customers raises apprehensions. OpenAI also delayed the public launch of its GPT-5.6 model at the request of the U.S. government, limiting access to a small group of vetted partners.
