Trump Lifts Block on Anthropic's Fable 5 and Mythos 5 AI to Rival China
US President Donald Trump has officially lifted the ban on the world’s most powerful artificial intelligence (AI) systems owned by Anthropic. This move follows the startup’s commitment to tighten security controls and work more closely with the federal government on testing models before they are released to the public.
This decision marks the end of a 19-day period of tension since the White House issued export control orders that effectively prohibited the use of the Fable 5 and Mythos 5 models. The lifting of the ban is viewed as a strategic move by the US to ensure that their technological advantage is not overtaken by the rapid progress of Chinese AI companies.
Anthropic has confirmed that access to the Fable 5 and Mythos 5 systems has been restored. Previously, US Department of Commerce export orders prohibited non-US citizens from accessing both systems. Because Anthropic did not collect user citizenship data, the company had taken the extreme step of blocking all access, including for US citizens themselves.
Fable 5 and Mythos 5 are categorised as a new class of super-powerful AI with the ability to detect and exploit software bugs. This capability is what triggered cybersecurity concerns in Washington.
Government intervention last June was prompted by findings from Amazon researchers regarding jailbreak methods capable of bypassing Anthropic’s security protocols. Although Anthropic CEO Dario Amodei initially downplayed the severity of these findings, the company eventually agreed to close the loopholes as a condition for lifting the ban.
However, Anthropic has acknowledged that these tightened security measures have impacted system capabilities. The company stated that their models are now less capable than previous versions because some user requests will be downgraded to weaker AI models to prevent cyber misuse.
The primary reason behind Donald Trump’s softening stance is the fear that strict restrictions would instead create space for China to lead. Recently, the Chinese company Zhipu AI released a system that researchers claim is capable of matching the capabilities of Mythos in finding security bugs.
In a letter addressed to Anthropic’s Chief Compute Officer, Tom Brown, US Secretary of Commerce Howard Lutnick emphasised that Anthropic has agreed to proactively detect security risks and report suspicious activities to the US government.
The relationship between Anthropic and the White House has not been harmonious. Dario Amodei has frequently criticised Trump’s leadership style, while Trump has previously referred to Anthropic’s management as a radical left-wing group. However, the necessity of winning the AI race appears to have forced both parties into a truce.
This move by Anthropic follows that of its main rival, OpenAI, which last week also agreed to limit access to its latest model, GPT 5.6, at the request of the US government. This indicates a new trend where giant AI companies are now under strict supervision by US defence and trade authorities.