Anthropic has dispatched senior technical staff to Washington as the artificial intelligence company seeks to resolve a dispute with the Trump administration that led to restrictions on its latest AI models and forced the company to disable access worldwide.
The meetings come amid growing government scrutiny of advanced AI systems and their potential use in cybersecurity, defense and national security.
Senior Anthropic technical staff are in Washington to meet White House officials following administration action against the company's Fable 5 and Mythos 5 models, according to Axios, which first reported the meetings Sunday, citing a source close to the company.
Anthropic staff have also participated in virtual discussions with White House officials since the administration first reached out on Friday, the outlet noted. Sources familiar with the situation told the publication that both sides are interested in finding a resolution.
The talks follow a Trump administration order barring foreign governments, companies and individuals from using Anthropic's Fable 5 and Mythos 5 models. As a result, Anthropic disabled access to the models globally to comply with the restrictions.
The San Francisco-based startup had previously limited access to its powerful Mythos model because of concerns about its cybersecurity capabilities. Earlier this week, Anthropic released Fable 5, a public-facing version of the technology that included what the company described as additional cybersecurity safeguards.
According to The Wall Street Journal, the dispute intensified after researchers at Amazon demonstrated ways some of Fable's safeguards could be bypassed to obtain information about software vulnerabilities. The findings prompted discussions among White House officials about how to respond to the model's capabilities.
The Journal reported that Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessent and National Cyber Director Sean Cairncross participated in discussions with Anthropic executives and technical leaders. Those discussions included Anthropic co-founder and Chief Compute Officer Tom Brown as well as Sarah Heck, the company's head of public policy.
Anthropic said Friday that the vulnerabilities identified by Amazon researchers were relatively basic and that comparable information could be obtained through other publicly available AI systems. Some cybersecurity experts have argued that the findings did not amount to a full jailbreak of the model and reflected capabilities that can also assist network defenders and software security teams.
The dispute has drawn attention from the cybersecurity community. On Sunday, a group of cybersecurity experts signed a letter urging the administration to reverse the restrictions, arguing that access to advanced AI systems can help identify and patch software vulnerabilities. The letter stated that the controls had removed important tools from defenders while creating uncertainty around U.S. AI leadership, according to The Wall Street Journal.
Anthropic, one of the leading U.S. AI developers, has positioned itself as a company focused heavily on safety and responsible AI deployment. The startup has confidentially filed for a U.S. initial public offering and has worked with government agencies and private-sector partners on AI safety and cybersecurity initiatives.
The administration has not publicly outlined any changes to the restrictions, while Anthropic has not disclosed details of the ongoing discussions beyond confirming compliance with the government's order.
