Monday, August 17, 2026

1. Signal of the Day · Regulation & Policy Surfaced late · published July 7

   FTC Proposes That State AI Laws Requiring Altered Model Output Are Preempted by Section 5

   The Federal Trade Commission's proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems (91 FR 41638, FR Doc 2026-13628) argues that AI companies which covertly steer model outputs away from users' reasonably expected objectives are likely to deceive consumers under Section 5 of the FTC Act — and that this holds "even if the deceptive steering is done in an effort to comply with State laws." Issued at the direction of Executive Order 14365 and announced July 1, 2026 on a 2-0 Commission vote per the FTC's press release, the statement names Colorado's AI Act (SB 24-205, repealed and reenacted as SB 26-189) and concludes such a law is impliedly preempted where it conflicts with the federal scheme. Comments closed July 31 with 40 received; no final statement has issued. Healthcare technology leaders who have anchored AI governance programs to state-law compliance should note that the state layer now has a live federal challenge attached, while the statement's disclosure standard — a disclaimer that "could not be buried in terms of service" — is a design requirement worth building to regardless of how preemption resolves.

   HighBy Federal RegisterRead at source ↗

2. AI Tech & Frameworks Surfaced late · published August 13

   Medicare Approves Add-On Payment for a Foundation-Model AI Device — Inpatient Only

   Aidoc announced on August 13 that CMS approved a New Technology Add-on Payment for CARE Body CT Multi-Triage, an FDA-cleared triage tool built on the company's diagnostic AI foundation model. Beginning October 1, 2026, qualifying Medicare Fee-For-Service inpatient cases are eligible for NTAP reimbursement over three years; the device received FDA Breakthrough Device Designation in September 2025 and clearance in January 2026. The scope guard matters as much as the news: the authority cited is the FY2027 Hospital Inpatient Prospective Payment System final rule (91 FR 49570). NTAP is an inpatient DRG add-on, and the Skilled Nursing Facility PPS is a separate payment system that is entirely unaffected — the release does not mention skilled nursing, post-acute or long-term care anywhere. Nothing here creates post-acute reimbursement for AI. The release also describes the decision as "novel," not "first" — a distinction trade coverage has begun to blur. What transfers is the precedent that Medicare will pay a premium for a foundation-model-based clinical tool, and the documentation burden that follows.

   MedBy PR NewswireRead at source ↗

3. SNF / Post-Acute

   Payroll-Based Journal Reporting Moves to iQIES Today; Data Accuracy Feeds Survey Eligibility

   Per CMS memo QSO-26-12-NH, the Fiscal Quarter 4 Payroll-Based Journal submission window opens August 17, 2026 in iQIES only; QIES stopped accepting PBJ records at 11:59 p.m. Eastern on August 14. Access is not automatic — submitters need a HARP account, an explicitly requested iQIES PBJ role, and approval from the facility's Provider Security Official, and vendors must be approved separately at each facility by CCN. The governance dimension is easy to miss: PBJ is auditable staffing data driving the Five-Star staffing rating, and "accurate data submission to CMS" is one of four qualifying criteria for the risk-based survey approach going national September 8 under QSO-26-14-NH. Skilled nursing and senior living operators layering AI on scheduling, agency-hour optimization or PBJ mapping face a system-of-record migration in which an automated pipeline can silently break or mis-map contract versus employee hours — surfacing months later as a quality-rating and survey-eligibility problem rather than an IT ticket.

   MedBy CMSRead at source ↗

4. Regulation & Policy Surfaced late · published August 3

   HHS Discloses Its Own LLM Use Inside a Final Rule — a Template for AI-Use Disclosure

   Buried in an HHS final rule on family violence prevention services (91 FR 48775, FR Doc 2026-15681, effective October 2, 2026) is a methodology note with governance significance well beyond its subject matter: "To support the analysis of public comments, HHS used a large language model, a type of artificial intelligence (AI), as a tool to conduct an initial scan of comment content, sentiment, and language. The AI output was thoroughly analyzed and refined by content experts. All comments were reviewed by OFVPS staff." That is a three-part human-oversight disclosure — tool named, role explicitly bounded to an initial scan, independent human review asserted as complete — published in binding form by the department that regulates the healthcare sector. The rule itself does not touch post-acute care and imposes no AI requirement. It matters as precedent: it is the clearest available model of how a federal regulator thinks AI-assisted analysis should be documented, and a better drafting template for an organization's own AI-use disclosures than most vendor governance material.

   MedBy Federal RegisterRead at source ↗

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