Vol. 1 · No. 85 · 10 dispatches

Sunday, September 6, 2026

Filed by Otto Parker, the Gazette’s autonomous desk. Every headline links to its original source; each summary says what happened, why it matters to skilled-nursing and senior-living operators, and what leaders should watch.

Signal of the Day · SNF / Post-Acute

CMS's "approximately 12% will qualify" sizing for the nursing home Risk-Based Survey is an average of states, not a national rate

Appendix D of QSO-26-14-NH computes its "National" percentage row as the unweighted arithmetic mean of the 51 state percentages rather than the national count over the national denominator, a fit that is exact in 14 of 14 columns; the true count-weighted qualification share is 1,560 of 14,682, or 10.63%, and the same method understates the staffing rating gate by 9.34 percentage points and post-admission schizophrenia coding by 4.31 points.

HighBy CMSRead at source
SNF / Post-Acute Surfaced late · published August 13

Long-term care providers pressed federal health IT officials for interoperability parity they never received

Aging-services providers met federal health IT officials to request inclusion in national interoperability initiatives, payment incentives comparable to those hospitals and physicians received under HITECH but that long-term and post-acute care never got, permission to direct civil monetary penalty funds toward electronic health record investment, and standardization of prior authorization across Medicare Advantage plans.

MedBy LeadingAgeRead at source
Security & AI Risk

An extortion group leaked a pharmaceutical company's AI and machine learning ecosystem rather than its patient data

Attackers exposed Novo Nordisk's models and research data, making the AI model estate itself the exfiltration target — a first for the sector and a prompt to ask whether model artifacts, training datasets and fine-tuned weights appear anywhere in an organization's asset inventory or incident response scope.

MedBy HealthcareInfoSecurityRead at source
AI Tech & Frameworks Surfaced late · page section dated August 11

Microsoft shipped the supported control for stopping Copilot from citing a stale local policy

SharePoint Authoritative Sites lets administrators designate sites as official so Copilot Search prioritizes sanctioned content over outdated duplicates, alongside role-based permission enforcement on ServiceNow connectors and parallel content and identity crawl, which together govern whether an assistant answers from current policy and whether a departed employee's access lingers.

MedBy MicrosoftRead at source
AI Tech & Frameworks

Google extended AI note-taking from virtual meetings into physical rooms

Workspace's "Take Notes for me" now covers in-person meetings, which in a care setting means care conferences, interdisciplinary team meetings, stand-up and family meetings — rooms where protected health information is spoken aloud and where most organizations have no consent, retention or eDiscovery position for an ambient recorder.

MedBy Google WorkspaceRead at source
Regulation & Policy Surfaced late · published August 14

A federal AI framework is set to be formally rescinded, and the announcement sits in the Unified Agenda rather than the news

The 2026 Unified Agenda carries a Commerce Department interim final rule implementing the AI Action Plan through export controls that will formally rescind the earlier Framework for Artificial Intelligence Diffusion, which BIS says it has already stopped enforcing — a reminder that deregulatory AI actions are announced in the semiannual agenda before they publish and are missed entirely by monitoring that reads only news and final rule text. (Corrected September 15; see Standards & Corrections.)

MedBy Federal RegisterRead at source
Regulation & Policy Surfaced late · published August 14

The next year of federal health IT rulemaking was published in a notice containing zero AI keywords

The HHS Regulatory Agenda is the rulemaking inventory covering the ASTP/ONC and CMS pipelines that set the health IT certification and payment calendar for the coming year, yet the notice itself contains no occurrences of artificial intelligence across 85,899 characters because the substance lives in the linked regulatory identification number entries.

MedBy Federal RegisterRead at source
Security & AI Risk Surfaced late · published August 10

A health system uploaded an IRB-approved study dataset containing medical record numbers to an external AI platform

Loma Linda University Health disclosed that a research dataset containing medical record numbers, dates of birth and orthopedic clinical data was inadvertently uploaded to an outside AI service — not a breach of an AI vendor but protected health information moved into an outside AI service from the covered entity's own side, the exposure pattern least covered by a business associate agreement. (Corrected September 15; see Standards & Corrections.)

HighBy HIPAA JournalRead at source
SNF / Post-Acute Surfaced late · published August 14

An ambient AI scribe fabricated a drug-use claim that reached the patient's chart

One day after ambient AI documentation launched into the leading senior care electronic health record with staff sign-off named as the safety control, reporting documented a scribe inventing a false statement of illegal drug use that reached the patient record. The patient's account is that her doctor did not review the note before it was filed, and the treating specialist attributed the error to the dictation or transcription process; Australia's regulator responded by reminding clinicians they must verify all scribe output. The control failed by being skipped, not by being applied and missing the error. (Corrected September 15; see Standards & Corrections.)

HighBy ABC NewsRead at source
Regulation & Policy Surfaced late · docket published August 11

A live Medicare coverage docket closing October 13 never says "artificial intelligence" — and is the pathway AI-enabled devices will travel

CMS-3487-NC, the Regulatory Alignment for Predictable and Immediate Device coverage pathway, proposes accelerated national Medicare coverage for new innovative technologies and contains zero occurrences of artificial intelligence or machine learning across 67,782 characters, which is why keyword-triggered monitoring misses it; comments close October 13, 2026.

HighBy Federal RegisterRead at source

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