Amazon security findings preceded Anthropic AI export limits
A Wall Street Journal report ties the restrictions on Anthropic’s Fable 5 and Mythos 5 models to Amazon research and CEO Andy Jassy’s talks with officials.
By James Whitfield · Staff Writer
3 min read
U.S. export controls that forced Anthropic to limit access to two AI models followed cybersecurity findings from Amazon and discussions between Amazon CEO Andy Jassy and the White House, The Wall Street Journal reported. The restrictions matter because they cut some foreign nationals off from Fable 5 and Mythos 5, including people tied to Anthropic’s own research work, according to The Verge.
The Journal reported that Amazon produced research claiming Fable 5 could be pushed, through a sequence of prompts, to provide information that might assist cyberattacks. The Verge reported that Amazon had not responded to its request for comment.
According to the Journal, Jassy shared Amazon’s findings with government officials shortly before the government moved to restrict use of the models by foreign nationals. Anthropic then cut access to Fable 5 and Mythos 5 after the government order, The Verge reported.
The order created an unusual problem inside Anthropic, according to The Verge. Many of the company’s researchers are foreign-born, and the restrictions meant some were blocked from using the company’s own technology.
Anthropic disputes the government’s framing
Anthropic challenged the government’s description of the security issue in a public statement. The company said officials had characterized the matter as a “jailbreak,” but Anthropic argued that similar weaknesses could be found through other public AI models, including GPT 5.5.
Some outside security voices echoed Anthropic’s position. Katie Moussouris, founder and CEO of LutaSecurity, wrote on BlueSky that she had seen the Amazon paper and that, in her view, “It’s not a jailbreak.”
Kate Koren, a former Commerce Department official, told the Journal that the White House’s antipathy toward Anthropic may have played a role in the decision. The Journal’s report did not establish that political tension was the sole basis for the directive.
A strained relationship with Washington
The restrictions land after a long-running dispute between Anthropic and the Trump administration, according to The Verge. The company has refused to allow its AI systems to be used for mass surveillance of Americans or to run lethal autonomous weapons, The Verge has reported.
In February, Trump directed federal agencies to stop using Anthropic’s AI, according to prior reporting by The Verge. Hours later, Defense Secretary Pete Hegseth labeled Anthropic a supply chain risk, The Verge reported.
The relationship had appeared to improve before the latest dispute. The Verge reported that the government and Anthropic had worked together to expand access to Mythos, suggesting some cooperation on cybersecurity uses before the new order.
The latest move now puts Anthropic back in conflict with federal officials over how powerful AI models should be handled when security concerns intersect with export controls. The Journal’s report places Amazon’s research and Jassy’s White House conversations at the center of the decision-making that preceded the restrictions.
This story draws on original reporting from The Verge.