US Government's Anthropic AI Model Ban Questioned by Experts
The U.S. government's export-control order, which compelled Anthropic to remove its Fable 5 and Mythos 5 AI models, was reportedly driven by "personality differences" rather than an AI jailbreak threat, according to TechCrunch's Zack Whittaker. Security experts assert that the alleged guardrail bypass did not justify the order and warn that this action establishes a troubling precedent, allowing the government to unilaterally disrupt American software products without court approval. This move could potentially erode trust in U.S. AI providers.
The U.S. government's export-control order, which mandated Anthropic to take its Fable 5 and Mythos 5 models offline, was reportedly not primarily about an AI jailbreak threat. TechCrunch's Zack Whittaker suggests the directive was influenced by "personality differences" between the AI company and the Trump administration.
Security experts have voiced concerns, stating that the reported guardrail bypass did not warrant the export order. They warn that this action sets a significant precedent, allowing the government to unilaterally disrupt American software products without judicial oversight, potentially undermining confidence in U.S. AI developers.
Cybersecurity veteran Katie Moussouris, founder of Luta Security, revealed in a blog post that Anthropic had shared a private paper with her detailing an alleged guardrail bypass in Fable 5, authored by security researchers at Amazon. Moussouris's analysis indicated that the bypass should not have triggered an export control. She noted that the described behavior cannot be meaningfully rectified and any attempt to do so would weaken the model's defensive capabilities.
Moussouris and numerous other security researchers and experts have since urged the Trump administration to revoke the export control order, calling the decision to remove advanced cybersecurity capabilities from U.S. network defenders "dangerous." They suggest that while past administrations have made broad decisions due to knowledge gaps, such as the overly broad 2010s export law on cybersecurity tools, the current directive appears retaliatory.
Justin Hendrix, editor of Tech Policy Press, commented that the Trump administration's action is likely to raise alarms internationally regarding the reliability of American AI for critical applications. This implies that AI companies in the U.S. might be perceived as operating under potential government interference. The Trump administration has not publicly confirmed the reasons behind its export control directive.
Hendrix further suggested a "cloud of suspicion that senior officials are picking favorites based on personal and political factors," indicating the government has set a precedent for significant control over the release of American-made software.
(Source: Slashdot)
