Legal AI Beat Morning Briefing
-
Anthropic’s Project Glasswing: Anthropic shares unreleased Claude Mythos Preview with rivals Microsoft and Google to collaboratively test AI-driven cyber vulnerabilities, prioritizing safety over immediate release.
-
Google’s Quantum Encryption Warning: Google flags 2029 as tipping point for quantum attacks cracking current encryption in apps like Signal/iMessage, urging immediate post-quantum upgrades for legal data handlers.
-
US Soft Law Shift: Voluntary AI standards gain binding force via courts, sidestepping EU-style mandates and reshaping compliance as “soft law” evolves into enforceable norms.
-
OpenAI Backs Illinois Liability Shield: Despite 90% public opposition, OpenAI supports bill exempting AI firms from liability for events killing 100+ or costing $1B, testing limits of innovation protections.
-
Anthropic Claude Code Leak: Exposed source code for terminal coding tool proliferates malware-laden GitHub forks, exposing developers to supply-chain risks and potential litigation.
-
EU AI Act Curbs Police Facial Rec: Prohibits real-time public facial recognition by law enforcement with narrow exceptions, leaving enforcement ambiguities that could sustain surveillance practices.
-
Birthright Citizenship Oral Args: Justices probe “closed set” of exceptions doctrine, with Barrett/Kavanaugh analogizing to gun rights, signaling potential narrowing of 14th Amendment interpretations.
-
West Coast Chatbot Regulations: Oregon/Washington impose minor-protection mandates; Washington’s requires deep design changes beyond disclosures, forcing AI firms to rethink consumer bot architectures.
(248 words)
