Odaily, citing The Wall Street Journal, reported that US government export controls and access restrictions on Anthropic’s Fable 5 / Mythos 5 models were driven in part by Amazon’s cybersecurity research, as well as communications between AWS CEO Andy Jassy and the White House. The case connects Anthropic’s model safety boundaries, security findings from a major cloud company, and regulatory action by the US government.
Amazon testing raised concerns over model output
According to the report, research submitted by Amazon found that, through a series of prompt-based tests, researchers were able to induce Fable 5 to produce sensitive information that could be used in cyberattacks. Those findings triggered safety concerns around the model. Andy Jassy later escalated the findings to the US government level, which led the White House to adopt further restrictions, including a ban on foreign users accessing the model.
The restrictions concern Anthropic’s Fable 5 / Mythos 5 models. The report links the regulatory response to Amazon’s cybersecurity work, showing how internal security testing by a company can enter government decision-making and affect access rules for AI models. The US measures were described as both export controls and access limitations.
Policy differences between Anthropic and the White House resurfaced
Former US Commerce Department official Kate Koren said the White House’s existing policy stance toward Anthropic may also have influenced the decision. She noted that Anthropic has had disagreements with the White House over AI safety boundaries, including its refusal to allow its models to be used for mass surveillance or lethal autonomous weapons systems.
The report also said that Anthropic and the White House had eased tensions earlier this year and expanded cooperation. However, the latest dispute over export controls and access restrictions for Fable 5 / Mythos 5 may again intensify tensions between the two sides. The central issue is not limited to one technical test; it involves model output risks, government security judgments, corporate policy positions, and the regulatory steps that followed.