David Sacks, the former White House AI and crypto advisor, stated on the June 21, 2026, episode of the 'All-In Podcast' that the White House developed serious national security concerns regarding Anthropic’s AI models. According to Sacks, officials viewed the company’s technology as a potential cyber weapon, largely due to internal reports from private companies regarding jailbreak vulnerabilities found in the 'Fable' model. Sacks claimed these security concerns were exacerbated by earlier statements from Anthropic CEO Dario Amodei, who allegedly described an earlier system as possessing significant offensive cyber potential.

The situation escalated into a breakdown of trust between the government and the company. Sacks asserted that when federal officials attempted to contact Anthropic to discuss the severity of the Fable vulnerabilities, Dario Amodei refused to cooperate. While these claims remain unconfirmed by Anthropic, they contributed to a broader policy shift within the administration. The Pentagon subsequently blacklisted Anthropic from military use, categorizing the firm as a supply chain risk. Despite this, the administration has explored ways to bypass restrictions for its 'Mythos' model, reflecting internal divisions on the matter.

In response to these developments, White House officials have drafted new national security AI policies. These directives mandate that federal agencies utilize multiple AI service providers to reduce dependency and place stricter limits on the influence of private contractors over military command structures. This shift coincides with the transition of senior AI policy advisor Sriram Krishnan, who stepped down in late June 2026 after focusing on expanding national data center infrastructure under President Donald Trump’s administration.

David Sacks, the former White House AI and crypto advisor, stated on the June 21, 2026, episode of the 'All-In Podcast' that the White House developed serious national security concerns regarding Anthropic’s AI models. According to Sacks, officials viewed the company’s technology as a potential cyber weapon, largely due to internal reports from private companies regarding jailbreak vulnerabilities found in the 'Fable' model. Sacks claimed these security concerns were exacerbated by earlier statements from Anthropic CEO Dario Amodei, who allegedly described an earlier system as possessing significant offensive cyber potential.

The situation escalated into a breakdown of trust between the government and the company. Sacks asserted that when federal officials attempted to contact Anthropic to discuss the severity of the Fable vulnerabilities, Dario Amodei refused to cooperate. While these claims remain unconfirmed by Anthropic, they contributed to a broader policy shift within the administration. The Pentagon subsequently blacklisted Anthropic from military use, categorizing the firm as a supply chain risk. Despite this, the administration has explored ways to bypass restrictions for its 'Mythos' model, reflecting internal divisions on the matter.

In response to these developments, White House officials have drafted new national security AI policies. These directives mandate that federal agencies utilize multiple AI service providers to reduce dependency and place stricter limits on the influence of private contractors over military command structures. This shift coincides with the transition of senior AI policy advisor Sriram Krishnan, who stepped down in late June 2026 after focusing on expanding national data center infrastructure under President Donald Trump’s administration.