White House's export limits on Anthropic linked to concerns about Chinese access (www.semafor.com)

The White House has implemented export controls on Anthropic's Mythos AI model due to concerns that it may have been accessed by a China-linked group. This decision restricts access to the model and its consumer version, Fable 5, solely to U.S. citizens, effectively barring foreign nationals, including Anthropic employees, from using these advanced AI tools. Anthropic had carefully deployed Mythos since its launch in April, limiting its availability to a select group to mitigate the risks of misuse, particularly in cybersecurity contexts. This action is significant for the AI/ML community as it underscores growing national security concerns regarding AI technology and its potential misuse by foreign adversaries, particularly in light of the model's capabilities to identify vulnerabilities in computer code. The situation was reportedly escalated when Amazon informed the White House about a potential "jailbreak" in Fable 5, which could allow unauthorized access and exploit its guardrails. While Anthropic maintains that it prohibits access from within China and the incident did not arise from previous tensions with the administration, the event highlights the critical intersection of AI safety, regulation, and international security in today's fast-evolving technology landscape.