OpenAI Rolls Out ‘Advanced’ Security Mode for At-Risk Accounts

A Wired security roundup covering multiple high-impact incidents and industry responses: OpenAI is rolling out an “Advanced Account Security” option for at-risk ChatGPT/Codex users; spyware/stalkerware exposed ~90,000 screenshots from a European celebrity’s phone; researchers found unauthorized access to Anthropic’s Mythos model; newly deciphered Fast16 sabotage malware (pre-dating Stuxnet) was reported; AI-assisted North Korean actors used large-language and multimodal tools to scale fraud and theft; the EU age‑verification app was shown hackable in minutes; and large data exposures and sales