The latest
The White House imposed restrictions on Anthropic’s advanced AI model Mythos after concerns emerged that a group linked to China may have obtained access to the system, according to a person familiar with the matter.
The Trump administration instructed Anthropic on Friday to limit access to Mythos and its consumer-facing version, Fable 5, to U.S. citizens only. Rather than comply with those conditions, the company chose to withdraw both models from the market.
Details
• Anthropic launched Mythos in April under strict access controls, limiting its use to a select group of companies focused on identifying software vulnerabilities.
• The company has said the model can discover advanced coding flaws that could be exploited by malicious actors, making uncontrolled access a significant security concern.
• It remains unclear how the White House learned of the issue, which organization may have accessed the model, or how access was obtained.
• U.S. officials fear that if Chinese entities gained access to Mythos, they could attempt to reverse-engineer the system and replicate its capabilities through a process known as distillation.
• David Sacks, President Trump’s AI adviser, said the administration received warnings that Fable 5’s safeguards could be bypassed through a jailbreak.
• Sacks claimed Anthropic CEO Dario Amodei downplayed the severity of the issue and declined to address it adequately after being notified.
• A source close to the White House said Amazon alerted the government to the jailbreak concerns and that CEO Andy Jassy discussed the issue with administration officials.
• Anthropic denied that White House discussions about export controls involved concerns over Chinese access to Mythos. The company noted that access to its products from within China is already blocked.
A growing dispute
The controversy marks the latest clash between Anthropic and the Trump administration.
The company has opposed administration-backed efforts affecting AI regulation at the state level and is also engaged in a legal dispute with the Pentagon over the use of its models in autonomous weapons systems.
Despite those tensions, Sacks said previous disagreements played no role in the decision to impose restrictions. He argued that the current issue could be resolved if Anthropic addressed the administration’s security concerns.
What to watch
The next key question is whether Anthropic will modify Mythos and Fable 5 to satisfy U.S. government requirements and relaunch them. The case could also become a major test of how Washington handles advanced AI systems that officials view as potential national security risks.
