Just weeks after blocking two of Anthropic’s newest AI systems over national security concerns, US regulators have reversed course, clearing the company to resume global access while placing new security commitments at the centre of the agreement.
Anthropic says it will begin restoring access to its most advanced artificial intelligence models after the US government lifted export restrictions that had temporarily halted their availability.
The company announced that access to Claude Fable 5 and Mythos 5 would resume on Wednesday following notification that the US Department of Commerce had withdrawn the restrictions imposed earlier this month.
The suspension, announced on 12 June, came shortly after both models were introduced. US officials had raised concerns that the technology could potentially be exploited by cybercriminals to identify and take advantage of weaknesses in computer systems.
According to a letter from US Commerce Secretary Howard Lutnick, Anthropic has now agreed to strengthen the way it monitors and responds to security threats associated with its AI models.
“Anthropic has agreed to proactively detect and address security risks associated with the models,” Lutnick wrote.
The agreement also requires the company to work more closely with US authorities on future AI releases and to report any signs that its technology is being used for malicious purposes.
While restrictions have been removed, the Commerce Department made clear the decision is not permanent. Officials said they retain the authority to impose new controls if future security concerns emerge.
The episode highlights the increasingly delicate balance governments are trying to strike between encouraging rapid AI innovation and preventing powerful models from being misused.
Fable 5 and Mythos 5 are Anthropic’s newest offerings on its Claude AI platform, which competes with products such as OpenAI’s ChatGPT and Google’s Gemini.
The consumer-focused Fable 5 is designed to tackle complex reasoning tasks and complete sophisticated assignments with greater independence. At the same time, Mythos 5 is aimed at enterprise users and cybersecurity professionals, offering advanced capabilities to identify software vulnerabilities and analyse computer code.
The company had previously questioned the government’s decision to suspend the models, saying officials had not identified any specific flaw that justified pulling the technology from users worldwide.
At the time, Anthropic said its understanding was that regulators were concerned about a possible technique for “jailbreaking” Fable 5, a method used to bypass built-in safety protections and unlock restricted behaviour.
The company argued that the reported issue represented only a limited security concern and maintained that it did not warrant withdrawing a commercial AI system already deployed to millions of users.
The decision to restore access suggests discussions between Anthropic and US regulators have resulted in additional safeguards rather than a permanent restriction, reflecting the growing scrutiny surrounding increasingly capable AI systems as governments seek stronger oversight without slowing technological progress.
