KPSGILL Weekly · Issue of · Permanent record — not overwritten by later issues
A federal AI docket opens, four regulators act on devices and medicines, and the consumer platforms move without a rule
Thirteen records at the close of business, 9:30 PM Pacific. FDA opened a docket on how generative-AI medical devices might be regulated and proposed nothing; the Medical Board of California meets 19 to 21 August with its materials published and no action yet taken; a Medi-Cal pharmacy policy takes effect 21 August; the Medicare drug-negotiation comment window closed 17 August without making the rule final; a second state joined the federal parity audits on prior authorisation; Australia’s radiopharmaceutical consultation closes 21 August; Europe reopened two medicine-development guidelines, one of them because medicines for quitting cigarettes were never studied in people dependent on vapes; the UK ordered devices out of use for missing certification rather than for any defect; Canada recalled pregabalin capsules contaminated with an antidepressant, an intravenous fluid and an airway device; and the consumer platforms shipped generative health features to hundreds of millions of phones without entering the pathway FDA is still designing. The distinction in each case is the record.
FDA opens a public docket on how it might regulate generative-AI medical devices
Jurisdiction
United States
Issuing body
FDA — Center for Devices and Radiological Health, Digital Health Center of Excellence
Instrument
Considerations for the Regulation of Generative AI-Enabled Medical Devices: Discussion Paper and Request for Feedback
State
Document state: DISCUSSION PAPER · Comment window: OPEN · Docket FDA-2026-N-7874
Dates
Issued · feedback due
Source read
The paper opens with a possible two-axis framework for assessing risk, then sets out a premarket approach built on competency assessment — explicitly modelled at a high level on how physicians are trained and evaluated — consisting of non-clinical device benchmarking followed by clinical confirmation that the device performs as intended before it reaches patients. It goes on to describe several risk-proportionate approaches to postmarket monitoring and raises considerations specific to foundation models and agentic systems. Each section poses targeted questions rather than answers.
For a physician the practical content is the frame, not a duty. If the competency model survives consultation, a generative tool arriving in a clinic will come with benchmark results and a clinical-confirmation record, which is a materially different purchasing and credentialling conversation from the one available now. FDA has put that shape in public before it has committed to it, and has left the mechanism — docket FDA-2026-N-7874 — open until 19 October.
What this does not establish. This is not a rule, not draft guidance and not final guidance. FDA states it is not proposing or implementing policy changes and is not communicating regulatory expectations, and the paper does not address whether the approaches it discusses fall within existing legal authorities or would require new ones. It imposes no obligation on any manufacturer, and none on a clinician using such a device.
The Medical Board of California meets 19–21 August with its materials published
Jurisdiction
California
Issuing body
Medical Board of California
Instrument
Noticed public meetings under the Bagley-Keene Open Meeting Act, Government Code section 11120 et seq.
State
Event state: UPCOMING · Materials state: PUBLISHED · Board action: NONE TAKEN
Dates
Panel meetings ; quarterly Board –
Source read
The Board's meetings index confirms quarterly Board and committee meetings running from the morning of 19 August to the close of business on 21 August 2026, with the agenda and supporting materials posted. Two state axes matter here and are frequently collapsed into one: the meeting is upcoming, and its materials are published. Neither makes anything on the agenda a decision.
Several agenda subjects would matter to practising physicians if the Board acts on them — legislation the Board tracks, its enforcement and investigation statistics, sunset-review proposals, continuing-education questions, the Interstate Medical Licensure Compact and a presentation on artificial intelligence in health care. This site's Medical Board dossier holds the pre-meeting record and will carry a dated action log after the meetings, appended rather than substituted.
What this does not establish. An agenda is not a Board decision, a staff recommendation is not Board action, draft minutes are not approved minutes, and a bill the Board supports is not law. Nothing on the published agenda has been adopted. This record asserts the meeting dates and the fact that materials are published; it asserts no figure, bill status or recommendation from the supporting documents.
The revised Medi-Cal Rx early-refill policy becomes operational on 21 August
Jurisdiction
California
Issuing body
Department of Health Care Services — Medi-Cal Rx
Instrument
Medi-Cal Rx early-refill policy for members aged 21 and older
State
Instrument state: FINAL · Operational from 21 August 2026
Dates
Effective
Source read
The change is operational rather than clinical: it governs when a pharmacy claim for a refill will pay and when it will reject, for members aged 21 and over. The consequence lands at the counter, on a Friday, which makes the following Monday the day a practice hears about it. The verified terms of the policy and its exclusions are held in the 17 August record and the Medi-Cal Change Ledger; this issue records the effective date, not a restatement of thresholds.
What this does not establish. This is a payment-policy change, not a change to what a physician may prescribe. It does not alter clinical judgement about quantity or timing, and it does not itself deny a medication; it determines whether a particular refill claim adjudicates on a particular date.
Two state axes move independently, and on 17 August only one of them moved. The comment window is closed. The instrument remains a proposed rule: nothing was finalised, no duty commenced, and no date of application was set by the window closing. The next event on this record is a final rule, whenever CMS issues one.
What remains open is elsewhere: CMS's draft guidance on manufacturer effectuation of the 2028 maximum fair price runs to 18 September, and the CY 2027 Physician Fee Schedule proposed rule runs to 14 September. Those are the windows where comment still changes anything.
What this does not establish. A closed comment period is not a final rule. It is not an effective date, and it creates no obligation. Nor does it foreclose the proposal: the rule stays PROPOSED until CMS acts.
Held with its primary source in the 17 August record, which preserves the pre-closure state.
United States · FDA · Device safety
Intraosseous needle sets recalled after reports the stylet can lock in place during insertion
Jurisdiction
United States
Issuing body
FDA — Center for Devices and Radiological Health
Instrument
FDA Early Alert, 4 August 2026; BD customer letter 30 July 2026; BD recall announcement 31 July 2026
State
Document state: EARLY ALERT · Firm action: RECALL TO USER LEVEL · Recall classification: not posted in FDA’s device-recall database at the date of reading
Dates
BD letter ; FDA early alert
Source read
Five powered-driver needle kits (15Ga, 15–55mm) were manufactured with out-of-tolerance dimensions, which can allow the obturator to rotate and lock into the needle hub during insertion. FDA reports that as of 22 July BD had reported 45 serious injuries and four deaths associated with the issue. Affected lots were distributed between 30 September 2024 and 12 June 2026; the powered driver itself is not affected.
The failure mode is confined to insertion, which shapes the response. FDA states that a needle already placed and functioning requires no further step and that a needle already used and removed requires no follow-up. Affected stock is to be identified and destroyed, not returned. FDA also makes the clinical point explicitly: an unsuccessful intraosseous attempt does not exhaust vascular access — further peripheral attempts or central access remain available on clinical judgement and local protocol. For emergency, critical-care and EMS settings the operational task is inventory, before the device is next needed in an arrest.
What this does not establish. An FDA Early Alert notifies the public of a potentially high-risk issue while FDA is still reviewing it. What the record establishes is an early alert plus a firm-initiated recall; a Class I designation for this 2026 event is not posted in FDA’s device-recall database at the date of reading and is not asserted here. A separate 2022 BD intraosseous recall was classified Class I and is a different event. Reported injury and death counts are BD reports to FDA as of 22 July, not adjudicated causation.
A second state audit lands on the same question: does parity reach prior authorisation?
Jurisdiction
United States — New York Medicaid programme
Issuing body
HHS Office of Inspector General
Instrument
Audit report A-02-24-01011 — New York Did Not Ensure That Selected Medicaid Managed Care Organizations Complied With Mental Health and Substance Use Disorder Parity Requirements Related to Prior Authorization
Action type
AUDIT FINDING — addressed to state oversight
State
Record state: CURRENT · Second report in the series after Kansas A-07-24-02842
Dates
Listed in OIG’s public report index; read
Source read
OIG has now put the same question to two states eight days apart. The Kansas report of 10 August and this New York report ask whether a state ensured that its Medicaid managed-care organisations applied mental-health and substance-use parity requirements to prior authorisation — not to the benefit package, which is where parity is usually argued and usually satisfied.
That distinction is the whole finding. A plan can carry behavioural-health benefits at nominal parity and still put a slower, narrower or more heavily reviewed authorisation pathway in front of them. Two states now have findings addressed to their oversight of exactly that pathway, which makes the authorisation record — timestamps, criteria applied, reviewer specialty, reason given — the document that carries an appeal, rather than the benefit schedule.
For patients. A denial of a mental-health or substance-use service is appealable, and the plan must give a specific reason for it. Nothing in this report changes a benefit or requires any patient to do anything. It does not establish that any particular denial was wrong.
What this does not establish. An audit finding is not a fraud finding, not a penalty, and not an adjudication against any plan’s clinical decisions; it is addressed to the state’s oversight of its contractors. This record establishes the report’s number, subject and jurisdiction as listed in OIG’s public report index on the date read, and dates itself to that reading. The report’s own cover date is not asserted here.
Public consultation: Radiopharmaceutical regulation in Australia, conducted through the TGA Consultation Hub
State
Consultation state: OPEN · Closes 21 August 2026
Dates
Closing date , 11:59 pm AEST · the information paper published with the consultation stated an original window of to
Source read
The subject is the regulation of medicines that carry small amounts of radioactive material and are used in the diagnosis and treatment of cancer and cardiovascular disease. Diagnostic and therapeutic products are both in scope, including theranostics. The TGA is asking how the existing settings work in practice — which pathway a hospital-prepared dose relies on, which exemption a nuclear-medicine service operates under, and how generator- and cyclotron-based supply is treated when models of care and manufacturing have moved.
The date matters more than the subject for anyone who wants a say: this is three days out, and it is a closing date rather than an opening one. It also carries a state change of its own. The TGA’s listing gives 21 August 2026; the consultation’s own information paper gives 5 June to 31 July 2026. The window moved out, and a reader who saw the earlier paper would otherwise conclude the consultation had closed three weeks ago.
For patients. This is a consultation about how a class of medicines is regulated in Australia. It does not withdraw, restrict or question any radiopharmaceutical, and no scan or treatment changes because of it.
What this does not establish. A consultation is not a regulatory decision. The TGA states that feedback from this consultation will not of itself produce immediate regulatory change and will inform future targeted consultation. No current obligation, supply pathway or exemption changes on 21 August; only the opportunity to comment ends. The reason for the extended closing date is not stated by the TGA and is not inferred here.
EMA moves to extend its smoking guideline to nicotine dependence that never involved smoke
Jurisdiction
European Union
Issuing body
European Medicines Agency
Instrument
Concept paper on the need for revision of the guideline on the development of medicinal products for the treatment of smoking, EMA/112569/2026
State
Document state: CONCEPT PAPER · Consultation: OPEN · Closes 30 August 2026
Dates
Consultation to · draft CHMP guidance planned no later than Q1 2027
Source read
EMA’s own statement of the problem is the useful part. The existing European guideline governs medicines developed for smoking cessation. It is neither intended for nor fully applicable to medicines for dependence on nicotine delivered without combustion — and because no other guideline applies, it is being used for those cases anyway. The proposal is to widen it to all forms of nicotine dependence, including inhaled and oromucosal products.
The clinical version of that gap is already in front of primary care. A nineteen-year-old dependent on a vape or a nicotine pouch, who has never smoked a cigarette, is treated with medicines whose authorisation rests on trials in smokers. This consultation is the regulator saying the evidence base was built for a different exposure, and it closes in twelve days.
For patients. Nothing changes about any treatment available now. It does mean European regulators have formally recognised that medicines for quitting cigarettes were not studied in people dependent on vapes or pouches — a fair question to raise with a clinician, and not a reason to stop a treatment that is working.
What this does not establish. A concept paper is not a guideline, not a decision and not a change to any authorisation. No development requirement moves, no indication is extended, and no product becomes approved or unapproved for nicotine dependence because of it. A draft revised guideline would itself go out for consultation before adoption.
The European biosimilar rulebook is reopened, and comparative efficacy studies are the question
Jurisdiction
European Union
Issuing body
European Medicines Agency — CHMP, Biosimilar Medicinal Products Working Party
Instrument
Concept paper on revision of the Guideline on similar biological medicinal products (CHMP/437/04 Rev. 1), EMA/CHMP/BMWP/153523/2026
State
Document state: CONCEPT PAPER · Consultation: OPEN · Closes 31 October 2026
Dates
Consultation to
Source read
The guideline being revised is the overarching one: the choice of reference product, how biosimilarity is established, and what clinical evidence a biosimilar developer is expected to bring. The live question is the necessity and utility of comparative efficacy studies, which EMA’s March 2026 reflection paper concluded do not add pivotal information where analytical and functional comparability is complete. The revision would also carry the new legal definitions from European pharmaceutical legislation into the guideline.
This is a prescribing question dressed as a methodology question. Biosimilar competition is what moves the price of a biologic and what makes switching pressure appear in a clinic; development cost and time are what determine whether a second or third biosimilar arrives at all. A guideline that stops expecting a comparative efficacy trial where the analytics are complete changes that arithmetic.
For patients. A biosimilar already prescribed in Europe is unaffected. This is about the evidence future biosimilars must bring, which over time affects how many exist and what they cost.
What this does not establish. A concept paper is not the revised guideline, and nothing in it removes an existing data requirement. It does not establish that comparative efficacy studies are no longer needed for any product, and it says nothing about the interchangeability or substitution of any authorised biosimilar. EMA has indicated a draft revision for consultation in 2027 and adoption in 2028; both are plans, not commitments.
The UK pulls devices from use for a paperwork failure, not a defect
Jurisdiction
United Kingdom
Issuing body
Medicines and Healthcare products Regulatory Agency
Instrument
Device Safety Information DSI/2026/007, under the UK Medical Device Regulations 2002; conformity also assessed against Regulation (EU) 2017/745
Action type
SAFETY ALERT — remove from use and quarantine
State
Record state: CURRENT · issued · no closing date; advice stands pending further MHRA direction
Source read
at gov.uk
The MHRA found medical devices on the UK market supplied without the conformity markings and certification that make supply lawful. Organisations are told to stop using and supplying the listed devices, retain them pending further advice, notify the agency at devices.compliance@mhra.gov.uk quoting DSI/2026/007, and help trace where stock went. Distributors are to cease supply and tell customers who may be holding it.
The device categories are the reason this matters more than its cause suggests: blood collection and infusion, surgical and biopsy procedures, wound management, and skin preparation and disinfection. One named product is a bone marrow biopsy needle. This is consumable stock that moves through a store room, not capital equipment on an asset register.
The agency is explicit that it has identified no defect, no performance issue, no quality issue and no safety signal. What is missing is the assurance — conformity assessment, quality-management systems, sterility, performance — that certification is supposed to establish before a device reaches a patient. There is nothing to inspect on the device itself, which is exactly why the remedy is an inventory check rather than a clinical one.
For patients. If you had a biopsy, a blood draw, an infusion or wound care in the UK this month, this alert does not say anything went wrong with your care and does not ask you to do anything. The MHRA has found no defect and no safety problem with the products. It is telling hospitals and suppliers to set aside stock whose paperwork does not prove it met UK standards.
What this does not establish. No patient harm, no device malfunction and no unsafe device is established. This is a Device Safety Information notice about regulatory status, not a recall of a defective product, and it makes no finding against any clinician or hospital that used the stock in good faith.
Canada recalls pregabalin capsules contaminated with an antidepressant
Jurisdiction
Canada
Issuing body
Health Canada
Instrument
Public advisory and health-product recall, Government of Canada Recalls and Safety Alerts database
Action type
RECALL — two lots
State
Record state: CURRENT · posted
Source read
Two lots of Teva-Pregabalin 150 mg capsules were recalled on 15 August for cross-contamination with sertraline — not a potency deviation or a particulate finding, but the presence of an entirely different medicine. A gabapentinoid contaminated with an SSRI.
That distinction changes the clinical question. The risk is not that the pregabalin dose is wrong; it is that a patient may be taking a serotonergic drug nobody prescribed. It matters most for someone already on an SSRI, an SNRI, a triptan or tramadol, and it matters in reverse too — an unexplained new symptom in a patient on pregabalin from these lots now has a mechanism worth asking about.
The practical task is identifying who holds the recalled lots, which is a dispensing-record question rather than a prescribing one. Pregabalin itself is not implicated and nothing about the medicine’s authorisation changes.
For patients. If you take pregabalin 150 mg in Canada, check with your pharmacy whether your capsules are from a recalled lot — the pharmacy can look this up from your dispensing record. Do not stop taking pregabalin on your own; stopping suddenly carries its own risks. Ask before you skip a dose.
What this does not establish. A recall establishes that specified lots must come out of supply. It does not establish that anyone was harmed, that every capsule in those lots is contaminated, or that either pregabalin or sertraline is unsafe as authorised. Lot numbers and quantities are held in Health Canada’s database entry and are not restated here.
Neither of these is a specialty product. Lactated Ringer’s is a standard intravenous fluid; a cuffed endotracheal tube is standard airway equipment. Both were posted to Canada’s central recalls register within the same week.
Particulate matter in an infusion and a defect in an airway device share a property that makes them worth pairing in one record: both are used at the moment there is least time to check a lot number. The remedy is therefore inventory-led — find affected stock before the next use, not at the next audit.
Health Canada’s recall rules were tightened in December 2024, when mandatory reporting timelines replaced expectations for Type I and Type II device recalls and the definition of a recall was widened to include corrective action the regulator orders. That is the machinery producing these postings on this cadence.
For patients. If you or a family member is in hospital in Canada, this needs nothing from you — recalled stock is identified and removed by the hospital and its supplier. Neither posting reports that a patient was hurt.
What this does not establish. These are recalls of specified products and lots. Neither establishes patient harm, and neither is a finding about a manufacturer’s conduct. Hazard classification and lot detail live in the database entries; this record does not assign a Type I, II or III classification that Health Canada has not published here.
The consumer health-AI layer moved while the device rules were still being written
Jurisdiction
United States · global consumer platforms
Issuing body
Company material — Google; Apple (reported, not announced)
Instrument
Google Health, successor to the Fitbit app (launched ), with a Gemini-based coach and a paid tier; release 5.05 on adding two-way Apple Health synchronisation and a link format for sharing data with providers
Action type
COMPANY MATERIAL — establishes only what the company says
State
Record state: CURRENT · no regulatory action, clearance or authorisation involved
Source read
Two things happened at opposite ends of the same question inside ten days. FDA opened a docket asking how generative-AI medical devices should be evaluated before and after market. The consumer platforms shipped generative health features to hundreds of millions of phones without entering that pathway at all.
Google’s app now summarises a user’s medical records and answers health questions conversationally, and the company states the coach is not designed to diagnose conditions or replace clinicians. Its 3 August release added synchronisation back into Apple Health and a link format for sending health data to a provider. Apple’s reported Health+ subscription with an AI agent has not been announced by Apple and is reporting, not company material.
The general-wellness framing is what keeps these features outside device regulation, and it is the same intended-purpose boundary the TGA regulates by and the FDA is now asking questions about. The boundary is the thing to watch, not the product. What arrives in the consulting room is more concrete: a patient with a platform-generated summary of their own records and a chatbot’s reading of it, raising two questions — whether the summary is complete enough to rely on, and who is accountable for an inference inside it.
FORECAST — LOW CONFIDENCE. An Apple Health+ subscription with an AI agent ships within twelve months. Basis: sustained reporting through 2025–2026 and a reported move of the health and fitness team under services. Competing scenario: Apple ships AI health features inside the existing free Health app with no subscription, or defers again — its Siri timeline has already slipped repeatedly, and Apple has announced neither product nor date.
For patients. A health app’s AI summary of your records is not a medical opinion, and the companies say so themselves. It can be useful for organising questions before an appointment. Two habits make it safer: bring the summary rather than the conclusion, and say where a claim came from, so your clinician can check it against the record.
What this does not establish. No regulatory action, clearance, authorisation or finding. No product named here is asserted to be a regulated medical device, and no claim about clinical performance is made or accepted. Google’s statement that its coach does not diagnose is recorded as the company’s position, not as a verified fact about the software. The Apple item is journalists’ reporting of unannounced plans.
The CHMP meeting of 17–20 August 2026 is under way. No opinion or outcome has been officially published, and none is reported here. Official highlights become a separate dated record after the meeting concludes.
Ten days in review
Every record dated to — 12 records across 5 jurisdictions, each at the authority that issued it. Records carrying earlier event dates sit in the issue where they were first published; the date below is the record’s own date, not the day it was noticed.
Swipe the table sideways for the remaining columns
12 records · 5 jurisdictions · 9–18 August 2026
Date
Jurisdiction & body
Record
Instrument
10 Aug
United Kingdom Medicines and Healthcare products Regulatory Agency
MHRA orders devices supplied without valid UKCA/CE certification out of use and into quarantine
SAFETY ALERT — DEVICE SAFETY INFORMATION
10 Aug
United States HHS Office of Inspector General
OIG finds Kansas did not ensure Medicaid managed-care parity compliance on prior authorisation
AUDIT / OVERSIGHT FINDING
11 Aug
United States Centers for Medicare & Medicaid Services
The full record store stood at 49 records at this issue’s close across six jurisdictions and is searchable from Search. Every record keeps its own address: 15 August, 16 August and 17 August remain exactly as published.
What changed state
A state change is not a new development. Each line below is the same record in a later condition; the earlier condition stays where it was published, on its own dated page.
Swipe the table sideways for the remaining columns
Unchanged and recorded as such: the intraosseous needle event remains an FDA early alert with a firm-initiated recall, and no recall classification for the 2026 event is posted in FDA’s device-recall database at the date of reading.
Fixed dates
Computed from the dated record store as at 18 August 2026, not typed into this page. A date that has passed moves to the closed rows with its transition; it is not removed.
Swipe the table sideways for the remaining columns
14 open dates and 1 recently closed, earliest first — 10 further fixed dates run to 2031 on the live desk
Every record in this issue sits one step short of the thing it is most likely to be mistaken for. The distinctions are the record.
A discussion paper is not guidance. FDA states its generative-AI paper proposes no policy and is neither draft nor final guidance. It does not even settle whether the approaches it raises fall within existing legal authority.
A closed comment window does not finalise a rule. CMS-4215-P remains a proposed rule. No duty commenced on 17 August and no date of application was set.
An agenda is not Board action. Nothing on the Medical Board’s published agenda has been adopted. A staff recommendation is not the Board’s position, draft minutes are not approved minutes, and the Board’s position on a bill is not that bill’s legal status.
An audit finding is not a fraud finding. The Kansas and New York parity reports are addressed to state oversight. They establish no violation by any plan, no penalty, and no finding about any clinical decision.
A consultation is not a regulatory decision. The TGA radiopharmaceutical consultation and both EMA concept papers invite comment. None of them changes an obligation, and a concept paper is not a guideline.
Accelerated approval is not traditional approval. The melanoma therapy recorded on 17 August rests on a surrogate endpoint and carries a confirmatory-trial obligation. Confirmed clinical benefit is not established.
A safety alert is not necessarily a recall, and a recall is not an injury. The intraosseous record holds an early alert and a firm-initiated recall together, with no classification posted. Reported injuries and deaths are reports to FDA, not adjudicated causation, and nothing in the record establishes harm to any individual patient.
A certification failure is not a device defect. The MHRA states it identified no defect, no performance issue, no quality issue and no safety signal in the devices it ordered quarantined. What is missing is the assurance certification provides, not evidence that anything went wrong.
A contamination recall is not a finding against the medicine. Two lots of pregabalin were recalled in Canada because sertraline was present in them. Pregabalin’s authorisation, safety profile and place in therapy are untouched, and no patient injury is established.
Company material is not a regulatory action. A health platform’s AI coach is not cleared, authorised or approved by anyone. That it sits outside device regulation is a consequence of its stated intended purpose — not a regulator’s judgement that it is safe or accurate.