US Restricts Frontier AI models

US Loosens Anthropic Claude Mythos Access, Unpatchable iPhone Exploit Emerges, and CISO Burnout Drives Fractional Shift Washington granted a partial reprieve allowing Anthropic's Claude Mythos to be released to more than 100 approved U.S. firms and institutions after export controls paused Mythos and the more restricted Fable 5, with access still limited to vetted American entities; the same day, OpenAI's GPT 5.6 was also restricted to government-approved partners under a Trump executive order requiring review of cyber-capable models.  The episode also covers Canadian hacktivist Aubrey Cottle's 18-month sentence for the 2021 Texas GOP hack and bail breaches, with possible U.S. charges pending. Researchers disclosed "USBliterate," an unpatchable physical USB exploit in the Secure ROM of older A12/A13 iPhones that aids forensic extraction. Finally, a survey finds rising CISO burnout, fewer full-time CISOs, growth in fractional CISO roles, and AI—especially shadow AI—overtaking liability as the top stressor. 00:55 AI Export Controls Shift 03:37 Anonymous Hacker Sentenced 05:32 Unpatchable iPhone Boot Exploit 07:30 CISO Burnout And Exodus 09:40 Wrap Up And Sign Off