The Trump administration has ended its temporary restrictions on Anthropic’s latest AI products, reopening access to the Claude Fable 5 chatbot for the public and permitting controlled use of its top-tier Mythos 5 model. The ban, implemented mid-June, had halted all usage following government intervention over cybersecurity risks.
Anthropic, an AI firm based in San Francisco, faced the block after its models were flagged for exposing software vulnerabilities. These concerns arose from findings by cybersecurity researchers at Amazon, Anthropic’s cloud services provider, who demonstrated a method to circumvent Fable 5’s safety measures and exploit potential security flaws.
Reflecting wider federal unease about the rapid advancement of AI technologies, officials had particularly focused on the Mythos model’s ability to scan for software defects that could be weaponized by hackers. Anthropic had previously alerted authorities to this capability, prompting heightened scrutiny from regulatory agencies.
This action is part of a broader federal effort under a recent executive order signed by President Donald Trump, establishing a framework for reviewing national security implications of cutting-edge AI tools before their public launch. While the order frames developer participation as voluntary, it has influenced companies like Anthropic and OpenAI to limit access temporarily.
In tandem with Anthropic’s situation, OpenAI also announced restrictions on its own new AI model, GPT-5.6 Sol, which will be available only to select government-approved users for the time being.