Anthropic executives are scheduled to meet with senior White House officials over national security issues related to their AI tools Fable 5 and Mythos 5, which have faced public access restrictions.
Executives from the artificial intelligence firm Anthropic are preparing for a meeting with senior officials from the White House, prompted by growing national security concerns regarding the company's latest AI tool releases. This meeting is scheduled for Monday in Washington D.C. and will involve discussions with the US Department of Commerce, which is overseen by Secretary Howard Lutnick.
The meeting follows Anthropic's decision on Friday to block public access to its newly released AI tool, which the company previously described as "too powerful." This action was taken after the US government prohibited the company from granting access to any foreign nationals.
The AI tool in question is known as Fable 5, which is a public version featuring additional safeguards, while Mythos 5, which has different controls, is restricted to a select group of organizations. Both tools are iterations of Claude Mythos, an Anthropic AI model that gained attention in April when the company initially allowed a limited number of organizations, including some government departments, to preview and test it.
Dario Amodei, the chief executive of Anthropic, is expected to attend the meeting alongside Secretary Lutnick, according to sources familiar with the situation. However, a spokesperson for the White House declined to provide any comments, and representatives from both the Department of Commerce and Anthropic did not respond to requests for further information.
Anthropic had previously indicated that releasing a public version of Mythos came with inherent risks, stating, "Fable's capabilities exceed those of any model we've ever made generally available."
Shortly after the release of Fable 5, the US government reported awareness of a potential "jailbreak" vulnerability, which refers to the possibility of manipulating the AI tool to perform unintended tasks. Anthropic clarified on Friday that it had only received verbal evidence regarding this alleged jailbreak issue.
The situation marks another chapter in the ongoing tensions between Anthropic and the US government, which have escalated this year, including a lawsuit by Anthropic against the Department of Defense concerning the use of its models. Despite these tensions, a prior meeting between Anthropic and senior White House officials had been characterized as "productive," suggesting a possible thaw in relations.
The upcoming meeting with the Department of Commerce is anticipated to involve more detailed documentation regarding the alleged jailbreak vulnerability, although it remains uncertain whether this will lead to the reinstatement of public access to Fable 5 and Mythos 5.
In a broader context, numerous leaders in the tech and cybersecurity sectors have urged the US government to permit Anthropic to release its models to the public. In an open letter, security personnel from companies such as Nvidia, Zoom, and Mercedes-Benz, along with former security staff from the US government and Google, appealed to Secretary Lutnick to lift the restrictions on Fable 5 and Mythos 5. They also called for the government to commit to a transparent and open process for addressing AI risk assessments moving forward.
Despite these calls for openness, the White House has indicated a preference for a relatively hands-off approach to AI regulation and has even expressed interest in the potential financial benefits of AI technologies. Nevertheless, the recent actions against Anthropic have sparked concern among AI developers and cybersecurity experts, with some arguing that restricting access to advanced capabilities could be detrimental, especially as adversaries continue to advance rapidly.
