Anthropic has pulled its Claude Fable 5 and Mythos 5 AI models after the U.S. government ordered the removal of safety guardrails [1].
This confrontation highlights a growing tension between private AI safety standards and national security requirements. The decision to disable the models suggests a fundamental disagreement over who controls the ethical boundaries of powerful artificial intelligence when federal interests are invoked.
The Trump administration reportedly demanded that Anthropic strip restrictions from the models to allow for unrestricted use [3]. This mandate was specifically tied to military and national security purposes [3]. According to reports, the government set a deadline of 5 p.m. [3] for the company to drop the restrictions.
Reports on the immediate outcome of the order vary. Some sources said the government successfully removed the restrictions from Claude Fable 5 and Mythos 5 [1]. However, other reports indicate that Anthropic responded by pulling or disabling the models entirely to prevent their unrestricted deployment [2].
Anthropic has not provided a detailed public statement on the specific technical nature of the guardrails in question. The company has previously emphasized the importance of safety frameworks to prevent the misuse of its large language models. The federal government said such restrictions hinder the capabilities required for defense operations [3].
The clash centers on the legal and ethical authority of the executive branch to compel a private company to alter the safety architecture of its proprietary software. While the government cites national security, the company appears to have prioritized its safety protocols over federal compliance in this instance [2].
“Anthropic has pulled its Claude Fable 5 and Mythos 5 AI models after the U.S. government ordered the removal of safety guardrails.”
This incident underscores a critical conflict between 'AI alignment'—the practice of ensuring AI behaves safely—and the strategic needs of the state. If the U.S. government can successfully mandate the removal of safety filters for military use, it establishes a precedent that national security overrides the safety charters of private AI labs. Conversely, Anthropic's decision to disable the models suggests that companies may prefer to kill a product rather than risk the deployment of an 'unaligned' or dangerous AI tool.
