Thursday, June 18, 2026

1. Signal of the Day · Security & AI Risk

   Anthropic Fable 5 and Mythos 5 Mandate 30-Day Data Retention Across All Platforms, Eliminating Zero-Retention Option

   Every prompt and response sent to Anthropic's Fable 5 and Mythos 5 models is retained for at least 30 days across all deployment surfaces—AWS Bedrock, Google Vertex, Azure Foundry, and Claude Enterprise—with zero-retention no longer available, creating a HIPAA-compatibility gap for healthcare organizations whose business associate agreements require zero-retention for PHI-adjacent workflows.

   HighBy PYMNTSRead at source ↗

2. AI Tech & Frameworks

   Microsoft Purview 9-Step Copilot Governance Deployment Now Mapped to HIPAA, NIST AI RMF, and EU AI Act

   A structured Purview-for-Copilot deployment framework covering sensitivity labels, data loss prevention, Insider Risk Management, Communication Compliance, six-year Copilot interaction log retention via Audit Premium, and Compliance Manager assessments pre-configured for HIPAA, SOC 2, NIST AI RMF, and EU AI Act gives healthcare IT organizations a tenant-side governance layer that operationalizes AI oversight controls within Microsoft 365.

   MedBy EPC GroupRead at source ↗

3. Regulation & Policy

   CMS WISeR Model Deploys AI to Screen Medicare Part B Prior Auth in Six States, Requiring Licensed Clinician Sign-Off on Denials

   The federal government's Wasteful and Inappropriate Service Reduction pilot—live since January 2026 across New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington—uses AI and machine learning to pre-screen 15 Part B services for prior authorization, but mandates licensed clinician review of every denial, codifying an 'AI assists, human decides' standard in Original Medicare.

   HighBy KFF Health NewsRead at source ↗

Get the Daily Dispatch by email · free, about 8 a.m. Eastern · RSS