Overview
- White House officials opposed Anthropic’s bid to add roughly 70 more preview users to its Claude Mythos cyber model, which would have lifted access to about 120 organizations, citing security risks and limited computing capacity.
- The Office of the National Cyber Director asked tech and security firms this week to detail how they will use AI to harden systems and prioritize fixes, with a Tuesday meeting outlining possible executive action to manage fast-rising cyber risks.
- Mythos is a restricted AI system that can find and chain software flaws across major operating systems and browsers, raising worries that discovery and exploitation could outpace patching if access scales too fast.
- Australia’s prudential regulator warned Thursday that banks are not keeping pace with AI-driven threats, and major Asian lenders said they are tightening checks as leaders at DBS, OCBC, UOB and Standard Chartered flagged faster, broader attack windows.
- Access remains gated under Project Glasswing for select partners such as Apple, Microsoft and Nvidia, while Anthropic investigates reports of unauthorized use, and OpenAI moved to a similar model with a limited rollout of GPT-5.5-Cyber for vetted defenders.