Monday, July 20, 2026

1. Signal of the Day · Regulation & Policy

   Senate votes 46–50 to preserve WISeR, keeping AI-assisted prior authorization in traditional Medicare

   The Senate defeated a Congressional Review Act resolution on July 16, preserving the six-year WISeR model that contracts private vendors to run AI-assisted prior authorization on select services in original Medicare across six states — the first affirmative congressional vote to keep algorithmic prior auth running in fee-for-service, a payer lane that historically required almost no pre-approvals.

   HighBy Emily Olsen · Healthcare DiveRead at source ↗

2. Regulation & Policy

   HTI-5 remains unfinalized, leaving proposed-for-deletion model card rules as the federal AI transparency floor

   Nearly five months after the comment period closed in February, the HTI-5 proposed rule has not been finalized, meaning the existing HTI-1 certification requirements for clinical decision support model cards technically remain in force even as the pending rule proposes removing them — the only federal AI transparency guardrail is the one slated for deletion.

   MedBy ASTP/ONCRead at source ↗

3. AI Tech & Frameworks

   Microsoft adds a governed request path for blocked AI agents in Microsoft 365

   Employees can now request access to blocked AI agents through the Agent Store with IT retaining final approval, and org-built agents require admin review before publication — converting unsanctioned AI use into an auditable request queue, alongside a new policy for watermarking AI-generated video and audio.

   MedBy Microsoft LearnRead at source ↗

4. SNF / Post-Acute

   Nursing homes turn AI on QAPI documentation gaps before they become survey deficiencies

   Skilled nursing operators are applying AI to quality assurance and performance improvement programs to surface documentation gaps ahead of surveys, denials, and audits, with the model framed as AI drafting and a clinician reviewing — proponents argue a clinician-reviewed AI-drafted discharge narrative is more likely to withstand audits and additional documentation requests.

   MedBy Skilled Nursing NewsRead at source ↗

5. Security & AI Risk

   Health-ISAC: conventional third-party risk review no longer covers AI vendors

   A July 14 Health-ISAC report argues healthcare organizations have little visibility into how vendor-embedded models are trained, updated, or subcontracted, and recommends AI-specific procurement policy, vendor model and training-data transparency, continuous lifecycle monitoring instead of point-in-time assessment, and explicit contract terms covering AI governance, incident reporting, model updates, and accountability.

   HighBy Pietje Kobus-McAllister · Healthcare InnovationRead at source ↗

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