Cyber leaders defend Anthropic's banned model (www.axios.com)

Prominent cybersecurity leaders, including executives from Adobe, Zoom, and Sophos, are urging the Trump administration to lift restrictions on Anthropic's advanced AI model, Mythos 5. They argue that limiting access to this technology could hamper cyber defenders just as they face an increase in AI-driven hacking threats. The group, led by former Facebook Chief Security Officer Alex Stamos, emphasizes that the vulnerabilities identified by Amazon researchers exist in other AI models as well, raising concerns about national security implications and market uncertainty in the AI landscape. The security experts assert that while the model can produce "proof of concept" vulnerabilities that aid both attackers and defenders, it's crucial for cybersecurity professionals to have access to these advanced tools. Stamos warns that restricting such models gives adversaries an advantage, particularly as non-Western AI technologies are rapidly advancing. He argues that shutting down capabilities like Mythos 5 during a time of increasing global cyber threats is not only imprudent but risks falling behind in the race to secure systems against vulnerabilities. The letter, which has gathered over 40 signatures, remains open for additional endorsements from other security leaders.