Unauthorised Access to Claude Mythos Prompts AI Security Concerns

In Other News: Unauthorized Mythos Access reports that Anthropic’s Claude Mythos was reportedly accessed by unauthorized users through a third‑party vendor environment, with Bloomberg News stating the interface was discovered and access subsequently restricted. The piece also notes that Sean Plankey has officially withdrawn his nomination for the role of Director at the Cybersecurity and Infrastructure Security Agency, leaving the agency without a permanent leader for now.

Meanwhile, the UK’s National Cyber Security Centre has unveiled a hardware guard called SilentGlass, designed to stop data exfiltration via high‑resolution display links and now available after government testing. The SecurityWeek roundup also highlights a joint FBI/CISA advisory about a large botnet of compromised SOHO routers and IoT devices controlled by China‑linked threat actors, under Volt Typhoon, with guidance issued to bolster defences.

Taken together, the items underscore ongoing concerns around unauthorised access to advanced AI capabilities, leadership vacancies in US cyber protection, and new mitigations for display‑link data leakage. according to Bloomberg News