How security concerns triggered WH showdown with Anthropic’s CEO? This is what we know

Anthropic’s decision to abruptly disable two of its most advanced artificial intelligence models just days after launch came after an intense behind-the-scenes effort by the White House to address what officials viewed as potential national security risks, according to a report by Politico. The report said the administration spent much of Friday (June 12) attempting to persuade Anthropic to voluntarily withdraw its newly launched Fable model after concerns emerged that its safety guardrails could be circumvented. When those efforts failed, the US government moved to impose export controls that effectively disabled Anthropic’s Fable 5 and Mythos 5 models only days after they became publicly available.

According to Politico, concerns about the models escalated on Thursday when Amazon chief executive Andy Jassy reportedly flagged issues related to bypassing the systems’ safety protections. Amazon, one of Anthropic’s largest investors, was responding to a request from the administration for feedback on the models, the report said, citing a person familiar with the discussions. Anthropic had launched the models with assurances that they incorporated strong safety measures. However, senior administration officials soon began questioning whether those safeguards were as robust as the company had claimed, Politico reported.

Within 24 hours, the matter had reached the highest levels of the White House. US Treasury Secretary Scott Bessent, White House cyber director Sean Cairncross, White House Chief of Staff Susie Wiles and other senior officials reportedly met to discuss the issue and determine an appropriate response. Following the meeting, officials attempted to contact Anthropic chief executive Dario Amodei. Administration officials cited by Politico claimed they were initially informed that Amodei was unavailable because he was attending a wellness retreat.

Anthropic strongly disputed that account. A company spokesperson told Politico that the claim was “absolutely false”. A person close to Anthropic told the publication that Amodei was first contacted around midday and joined discussions with senior administration officials within approximately 75 minutes. During the period he was unavailable, Anthropic had offered alternative senior executives to engage with the administration, the person added.

3 calls and a disagreement

Once communication was established, Amodei participated in three separate calls with senior administration officials, including Bessent, Cairncross and Commerce Secretary Howard Lutnick, according to Politico. During those discussions, Amodei reportedly argued that the administration’s concerns were based on a misunderstanding of the technical issue involved. The Anthropic chief executive maintained that the vulnerability being discussed differed from a broader "jailbreak" scenario that would allow users to bypass all of the company’s safety restrictions, the report said.

Despite those assurances, officials remained concerned that the models could be exploited in ways that posed national security risks. The disagreement ultimately proved irreconcilable, leading the administration to impose export controls rather than rely on a voluntary withdrawal by the company.