Amazon CEO Implicated in Global Suspension of Anthropic's Fable AI Model

Reports have surfaced indicating that Amazon's chief executive played a significant role in the recent global suspension of Anthropic's latest artificial intelligence model, Fable. According to multiple news outlets, the CEO of Amazon, Andrew Jassy, informed the United States government about a potential security vulnerability found in Fable, which ultimately led to the enforcement of stringent export restrictions and the model's subsequent lockdown.

Initially, Amazon's cybersecurity division conducted internal investigations into Fable, probing the model for weaknesses that could be exploited through so-called jailbreak techniques. Jailbreaking, in this context, refers to the process of guiding an AI system to circumvent its built-in safety protocols and execute prompts that would normally be blocked for security reasons. The internal tests reportedly demonstrated that Fable could generate working exploits for vulnerabilities in four widely-used software programs. These findings were reportedly escalated by Amazon's leadership to officials within the US government, including Treasury Secretary Scott Bessent.

While Amazon has not commented on the specifics of the case, a company spokesperson confirmed that government agencies frequently consult with Amazon on technology-related security concerns. In recent years, Amazon has invested heavily in the AI sector, pledging $50 billion in funding to OpenAI and committing an additional $25 billion over the coming years to Anthropic, further highlighting its strategic interest in generative AI advancements.

Following Amazon's report, the US government delivered an ultimatum to Anthropic: address the identified security issues associated with jailbreaks within 24 hours or restrict access to Fable for non-US users. David Sacks, a technology investor and advisor to the US President's Council on Science and Technology, detailed the sequence of events on social media. According to Sacks, Anthropic's CEO, Dario Amodei, declined to resolve the reported vulnerabilities, maintaining that the incidents did not constitute a full jailbreak and that similar outputs could be achieved with competing models such as OpenAI's ChatGPT 5.5. Nevertheless, the US government proceeded with the directive, resulting in the worldwide suspension of Fable for users outside the United States. This measure also affected many foreign employees of Anthropic, who are now unable to access or develop the model.

The US government's approach has drawn scrutiny, as the decision to suspend a single AI model due to potential jailbreak vulnerabilities raises questions about the broader implications for the industry. AI experts note that, in theory, any AI model is susceptible to jailbreak attempts, making targeted restrictions potentially inconsistent. Fable, a variant of Anthropic's earlier model called Mythos, had already attracted attention for its capacity to identify and exploit novel software vulnerabilities rapidly, prompting Anthropic to limit its public release due to safety concerns.

This latest intervention is seen against the backdrop of ongoing tensions between Anthropic and US regulatory authorities. Previously, the US government imposed sanctions on Anthropic after the company refused to permit the use of its AI technology in autonomous weapons systems, citing national security and supply chain risks. Anthropic has challenged these measures in court. Whether similar legal actions will arise from the Fable suspension remains to be seen.

Sources close to the US government have indicated that it is unlikely other AI developers will face comparable restrictions solely based on jailbreak potential. Nevertheless, the episode has sparked debate within the technology sector regarding the appropriate balance between innovation, security, and regulatory oversight, particularly as AI systems become increasingly sophisticated and impactful.