Days after the U.S. government prohibited access to Claude Fable 5 and Claude Mythos 5 to all non-American nationals including foreign employees of Anthropic, David Sacks, chair of president’s council of advisors on science and technology said that it was the “jailbreaking” or lack of safety measures in Fable AI model that led to the ban.
Anthropic had released Claude Fable 5, a “safe” version of its frontier AI model Claude Mythos, for general public use earlier June, claiming that the AI model comes with necessary guardrails to avoid its misuse. However, as per Sacks, it was a tech giant that tested Fable 5 model and found “jailbreaking” techniques to avoid the safety guardrails.
Jailbreaking in AI refers to ways to manipulate an AI system or an LLM to bypass its safety checks and ethical guardrails.
Read More: What is AI Jailbreaking?
Taking to X on Saturday, Sacks accused Anthropic of not willing to fix the jailbreaking issue and rather prioritizing the “continued offering of the consumer model over safety.”
Without naming the tech giant, Sacks informed that a “highly credible, trusted partner” of both Anthropic and the US government had tested the Fable 5 model and found a way to bypass guardrails. According to an exclusive report by Wall Street Journal, it was Amazon CEO’s talks with U.S. officials regarding the Fable AI model that led to the ban.
“A highly credible trusted partner of both Anthropic and the USG who was testing Fable came forward with a jailbreak of those guardrails. The Admin asked Dario (CEO of Anthropic) to fix the jailbreak or de-deploy the model. Dario refused.” wrote Sacks.
I’ve had a number of conversations with folks inside and outside government about the current situation with Anthropic, and here is what I believe to be true:
— As we know, Anthropic publicly released its Mythos class models earlier this week under the commercial name Fable.…
— David Sacks (@DavidSacks) June 13, 2026
Amazon is one of the several partners of Anthropic under Project Glasswing where the latter had shared exclusive access to Claude Mythos preview model for the former to find vulnerabilities in their system.
Meanwhile, Anthropic, in a detailed statement, has asserted that the capabilities demonstrated in the jailbreak and accessible through other public AI models did not warrant the removal of a widely used commercial product from the market.
Sacks mentioned that the federal government requested that Anthropic either close this loophole or temporarily remove the system from circulation until a resolution could be achieved. He asserted that Anthropic declined the government’s request, and, as a result, the federal government placed an export restriction on access to Fable 5 and Mythos by foreign nationals. Complying, Anthropic pulled the plug on Fable and Mythos models for all its customers and even employees.
Sacks stated that members of the administration were reluctant to impose the export restriction, and that they were surprised that Anthropic did not comply with the request to address a safety concern, given Anthropic’s long-standing commitment to creating safe AI systems.
Anthropic has disputed Sacks’ assertions regarding their interactions with the federal government and provided a response to the U.S. Department of Commerce stating that the company investigated reports of the jailbreak and found that the vast majority of capabilities demonstrated in the reported event were previously known to Anthropic.
Meanwhile Sacks claimed that the ball is in Anthropic’s court to decide if they wish to fix the guardrails so that the ban is lifted on Claude Fable.
“The Admin’s hope now is that Anthropic remediates the safety issue, the export control is lifted, and Fable goes back into general release. The Admin wants all of this to happen as soon as possible. It is frankly bewildered that Anthropic hasn’t wanted to comply with safety requests that it previously said were its highest priority,” added Sacks.
Also Read: Anthropic Has a History of Scare Tactics: David Sacks on Claude Mythos
