The Amazon research alleged that Fable 5 could be manipulated to provide instructions for cyberattacks, a claim that prompted immediate federal intervention. Anthropic, however, rejects the government's classification of these findings as a "jailbreak." The company contends that the identified vulnerabilities are common across the industry and exist in competing tools like GPT 5.5, a position supported by independent security experts like LutaSecurity’s Katie Moussouris.
Amazon Research and the White House Ban on Anthropic AI
A series of security prompts developed by Amazon served as the catalyst for the White House to restrict access to Anthropic’s Fable 5 and Mythos 5 models. Following a direct briefing from CEO Andy Jassy, the administration enforced export controls that inadvertently locked Anthropic’s own foreign-born researchers out of their systems.
This dispute marks another chapter in the strained relationship between the Trump administration and Anthropic. Tensions have simmered over the company’s public refusal to support mass surveillance initiatives or lethal autonomous weapons technology. Despite a brief period of cooperation to expand access to Mythos, the administration’s earlier designation of Anthropic as a supply chain risk, spearheaded by Secretary of Defense Pete Hegseth, suggests that the current export restrictions may be as much about political friction as they are about technical security concerns.
Comments (0)
No comments yet. Be the first!