Anthropic holds talks with US in bid to lift curbs on AI models
Anthropic met with US Commerce Department officials to address a Trump administration order halting foreign access to its advanced AI models, Fable 5 and Mythos 5, over jailbreak vulnerabilities. The company argues the restrictions could stall AI development across the industry, while officials seek to resolve security concerns with voluntary reviews and guardrail improvements.
Anthropic PBC held talks with US Commerce Department officials on Monday to resolve a dispute over national security concerns related to its AI models. The Trump administration had ordered the company to disable foreign access to its cutting-edge Fable 5 and Mythos 5 models, citing potential jailbreak risks that bypass safety guardrails. The directive, issued late Friday, marks the US government’s most direct intervention in AI operations to date. Anthropic complied by disabling access, arguing the vulnerability was narrow and that broader restrictions could halt AI advancements. The company stated that if applied industry-wide, such measures would freeze new model deployments for all providers. Anthropic’s spokesperson confirmed ongoing virtual meetings with US officials to address specific security issues. The company had previously restricted its Mythos model due to cybersecurity risks, releasing Fable 5 as a public-facing version with guardrails to limit its capabilities. However, the administration’s order introduced uncertainty for other AI firms and allied governments navigating compliance. Cornell University’s AI Strategy and Innovation director, Ayham Boucher, warned that framing the issue as a simple guardrail fix oversimplifies the technical challenges. He noted that jailbreak resistance is complex, unlike patching standard software bugs. Meanwhile, the Trump administration’s recent executive order on AI emphasized voluntary reviews without mandating licensing, adding to the regulatory ambiguity. The Commerce Department’s actions follow a broader push by the Trump administration to accelerate AI policy, with the order serving as a wake-up call for the industry. Joelle Pineau, Chief AI Officer at Canada-based Cohere Inc., highlighted the uncertainty created by the dispute, signaling potential ripple effects for global AI development.
This content was automatically generated and/or translated by AI. It may contain inaccuracies. Please refer to the original sources for verification.