San Francisco (United States) (AFP) – Anthropic will soon begin restoring access globally to its most powerful AI models, Fable 5 and Mythos 5, after the US government lifted a restriction on where they could be released, the company said Tuesday. Over the past couple of weeks, the Trump administration has invoked national security concerns to limit the ability of major US tech companies to release advanced models, including those from Anthropic, which some researchers feared could be exploited to bypass cybersecurity measures.
“We’ve received notice that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” Anthropic posted on X. “We’ll begin restoring access tomorrow.” Just four days ago, the company said it had received authorization from the government to allow a small group of American cybersecurity firms to access Mythos 5. Secretary of Commerce Howard Lutnick said in a June 26 letter to the company that “Anthropic has worked with the US government to address risks associated with the Covered Models,” Politico reported.
The government abruptly forced Anthropic to cut off access to its two cutting-edge artificial intelligence models on June 12 after discovering vulnerabilities in the safeguards put in place to prevent misuse of the tool. On Tuesday, Lutnick told Anthropic in a letter that the Trump administration had “withdrawn” its previous restrictions on the release of the company’s models, Politico reported. The letter indicated that the Trump administration was satisfied, at least for now, that Anthropic had “taken steps in close coordination with the U.S. government to address the risks associated with Claude Mythos 5 and Claude Fable 5.”
Like Anthropic, rival AI lab OpenAI has also complied with Washington’s requests to restrict its own release of a new, powerful model called GPT-5.6 to a limited set of approved partners. “This isn’t quite the process that we think is optimal,” OpenAI CEO Sam Altman said Friday in a post on X, alongside an explanation of the GPT-5.6 launch. Anthropic did not immediately respond to a request for comment Tuesday. However, in a blog post published Tuesday evening, Anthropic called for the development of a standardized “framework” to both assess critical vulnerabilities in advanced models and respond to them. The San Francisco-based AI lab will work with Amazon, Microsoft, Google, and others on the effort.
“This problem will become more acute in the coming months, as more models with powerful cybersecurity (and other) capabilities are trained, assessed, and released,” the blog post said.
– New frontiers –
The Trump administration issued an executive order on June 2 calling for the federal government to take multiple steps over the subsequent two months to take actions on AI and cybersecurity — including creating a voluntary “framework” for private companies, such as Anthropic and OpenAI, to test and release their powerful “frontier” AI models in collaboration with the government. Susie Wiles, the president’s chief of staff, posted Tuesday on X that the Trump administration was grateful for the cooperation from tech companies, though she didn’t name any.
“My gratitude to companies across industries who continue to work closely with the White House to implement the President’s executive order on AI and cybersecurity,” Wiles said. “This includes excellent work around advanced model access and guardrail testing and security.”
Earlier in the day, CIA Director John Ratcliffe compared the capabilities of the most advanced artificial intelligence models to nuclear weapons, in a tacit defense of the Trump administration’s recent hard line on controlling the release of the most powerful AI technology. “In conversations with many of the president’s other national security and economic security advisors, we’re talking about the impact of these frontier AI models,” Ratcliffe said during a speech at the AWS summit in Washington. “It would be…not misplaced to refer to their capabilities as akin to digital nuclear weapons,” Ratcliffe said.
© 2024 AFP
