Anthropic’s Fable 5 AI Model Taken Down After Amazon Security Report, White House Intervention

Anthropic’s Fable 5 AI model was abruptly removed from public access on Friday after Amazon alerted U.S. officials to potential national security risks, according to multiple sources. The shutdown followed hours of urgent negotiations between the White House and the AI startup, with the administration citing “national security threats” linked to Amazon’s findings.

Why Was Fable 5 Taken Down?

Amazon reported to the U.S. government that researchers had bypassed security measures in Anthropic’s Mythos AI model, a precursor to Fable 5, according to a source familiar with the matter. The tech giant shared the findings with administration officials on Thursday night, triggering a rapid response. By Friday evening, Anthropic had removed Fable 5 from public access, with the White House issuing a letter imposing export controls on the model.

“As a leading cloud provider, we don’t share details of these discussions,” an Amazon spokesperson said in a statement. The company did not confirm the specific security concerns but emphasized its role in advising clients on risks.

What Do Anthropic and Experts Say?

Anthropic disputed the severity of the reported threat, arguing that the security vulnerability identified by Amazon was “relatively simple” and not indicative of a flaw in Fable 5’s safety systems. The company’s CEO, Dario Amodei, and other executives engaged in multiple calls with administration officials, according to a source. However, the government maintained its stance, providing no further details on the alleged threat.

Independent security researcher Katie Moussouris, who reviewed the Amazon report for Anthropic, criticized the government’s response. “The administration’s actions seem way out of line with the actual findings,” she told Axios. Moussouris added that the researchers had identified vulnerabilities by asking standard questions, a process designed to test AI defenses.

What Are the Implications?

The export control letter, issued by the White House, restricts access to Anthropic’s advanced models for foreign nationals and U.S. allies. This move has raised concerns among the company’s international employees, many of whom rely on the models for their work. The rules also set a precedent for how the government might regulate AI models deemed “national security threats.”

“This is a de-facto licensing regime,” a source familiar with the administration’s thinking told Axios. The government has not indicated plans to apply similar restrictions to other AI models, according to an official, but the decision highlights growing scrutiny of advanced AI systems.

What Happens Next?

Anthropic has not yet commented on whether it will challenge the export controls or seek to re-release Fable 5. The company had previously notified the government about the planned June 9 release of Fable 5, with no objections raised at the time, according to a source. The administration’s sudden intervention has sparked debates about the balance between AI innovation and national security.

“There was a lack of seriousness in how Anthropic approached the release,” the source said, adding that the company’s “overly confident” stance may have contributed to the fallout. The incident underscores the tension between private tech firms and regulators as AI technology evolves rapidly.