Amazon CEO Andy Jassy privately raised national security concerns about Anthropic’s most advanced AI models to senior Trump administration officials before the government ordered the company to suspend global access to Fable 5 and Mythos 5, according to The Information. Reuters independently confirmed the outreach, reporting that other tech leaders also participated.
The behind-the-scenes lobbying preceded the June 12 export control directive that forced Anthropic to disable Fable 5 and Mythos 5 for all users globally. In a blog post, Anthropic said the government cited national security authorities and concerns about a potential jailbreak technique. Anthropic contested the move, arguing the identified jailbreak was narrow and non-universal, and that its safeguards were “substantially more effective than those of any previously deployed model.”
Amazon’s Position as Investor and Competitor
The disclosure complicates Amazon’s relationship with Anthropic. Amazon Web Services is Anthropic’s largest cloud partner and Amazon has invested up to $8 billion in the company. At the same time, AWS competes directly with Anthropic through its own Bedrock AI platform. Jassy raising security concerns to the government about a company Amazon simultaneously backs financially puts the cloud provider in an unusual dual role: investor and de facto regulator.
For teams building agent workflows on Claude-family models through AWS Bedrock, the episode raises questions about infrastructure provider neutrality. If a cloud provider can influence which models the government restricts, the risk calculus for choosing a model vendor now includes the political dynamics between that vendor and its cloud host.
The Government’s Rationale
According to Al Jazeera, the directive was issued partly over suspicions that a China-linked group had accessed the new models. The order bars all foreign nationals from accessing Fable 5 and Mythos 5, including Anthropic’s own foreign national employees.
Anthropic’s response was pointed: the company said it received the directive at 5:21 PM ET and was only shown a verbal demonstration of a “narrow, non-universal jailbreak” involving asking the model to read a specific codebase and flag software flaws. Anthropic noted that thousands of hours of red-teaming with the US government, UK AISI, and private organizations preceded the launch, and that no testers had found a universal jailbreak.
Cloud Providers as Policy Gatekeepers
The Amazon disclosure is one of the clearest examples of cloud infrastructure providers directly influencing government AI policy. AWS, Google Cloud, and Microsoft Azure collectively host the majority of frontier model inference. When those providers lobby governments about model safety, they shape which capabilities reach agent builders and which get restricted.
For organizations deploying autonomous agents that depend on specific model families, the takeaway is that model availability is no longer purely a technical or commercial question. It is also a political one, influenced by the competitive dynamics between the companies that build models and the companies that host them.