commit archived sources from previous research sessions

inbox/archive/health/2020-03-17-pnas-us-life-expectancy-stalls-cvd-not-drug-deaths.md

@@ -0,0 +1,38 @@
+---
+type: source
+title: "US Life Expectancy Stalls Due to Cardiovascular Disease, Not Drug Deaths"
+author: "Shiels MS, Chernyavskiy P, Anderson WF, et al. (NCI)"
+url: https://www.pnas.org/doi/10.1073/pnas.1920391117
+date: 2020-03-17
+domain: health
+secondary_domains: []
+format: research-paper
+status: unprocessed
+priority: high
+tags: [cardiovascular-disease, life-expectancy, opioids, drug-deaths, 2010-period-effect, mechanism, belief-1]
+---
+
+## Content
+
+Published in *PNAS*, March 17, 2020. NCI researchers. This is the foundational paper establishing that CVD stagnation — not drug deaths — is the primary driver of US life expectancy plateau.
+
+**Key findings:**
+- CVD stagnation held back US life expectancy at age 25 by **1.14 years in both women and men** between 2010 and 2017.
+- Rising drug-related deaths had a much smaller effect: **0.1 years in women and 0.4 years in men.**
+- Ratio: CVD stagnation effect is approximately 3–11x larger than drug mortality effect on life expectancy.
+- The stagnating decline in CVD mortality was "the main culprit outpacing and overshadowing the effects of all other causes of death."
+
+Context: This paper was published before the 2026 PNAS cohort analysis but establishes the primary mechanism. The 2026 cohort paper (Abrams & Bramajo) extends this finding by showing the same CVD-driven pattern operates at the cohort level with a distinct 2010 period effect.
+
+## Agent Notes
+**Why this matters:** This is the key mechanism paper for the disconfirmation search. The opioid epidemic was the popular narrative for US mortality stagnation; this paper shows CVD is 3-11x more impactful. Since CVD/metabolic decline is structural (not reversible like opioid epidemic), this STRENGTHENS Belief 1's "binding constraint" framing.
+**What surprised me:** The magnitude of the ratio — CVD effect is 3-11x drug deaths effect. Most public discourse attributes the stall to opioids. The actual driver (CVD/metabolic) gets far less attention.
+**What I expected but didn't find:** Opioid mortality being the primary driver. The data contradicts the popular narrative.
+**KB connections:** Directly relevant to any claim about structural health deterioration; connects to "deaths of despair" claims; links to food industry and metabolic disease claims.
+**Extraction hints:** "US life expectancy stagnation is driven primarily by CVD plateau (1.14 years lost), not drug deaths (0.1-0.4 years lost) — a 3-11x difference that inverts the dominant public narrative."
+**Context:** Published 2020, now confirmed and extended by 2025-2026 literature. The 2010 CVD stagnation pattern was visible even in 2020 data. This is not a new phenomenon — it's been building for 15 years.
+
+## Curator Notes
+PRIMARY CONNECTION: PNAS 2026 Abrams-Bramajo cohort paper (already archived); provides mechanism for 2010 period effect
+WHY ARCHIVED: Foundational mechanism paper establishing CVD>drugs as life expectancy driver; frequently cited in subsequent literature
+EXTRACTION HINT: Quantitative claim: "CVD stagnation costs 1.14 life expectancy years vs. 0.4 years for drug deaths — inverting the public narrative about opioids as the health crisis driver."

inbox/archive/health/2024-12-02-jama-network-open-global-healthspan-lifespan-gaps-183-who-states.md

@@ -0,0 +1,40 @@
+---
+type: source
+title: "Global Healthspan-Lifespan Gaps Among 183 World Health Organization Member States"
+author: "Garmany et al. (Mayo Clinic)"
+url: https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2827753
+date: 2024-12-02
+domain: health
+secondary_domains: []
+format: research-paper
+status: unprocessed
+priority: high
+tags: [healthspan, lifespan, disability-adjusted, WHO, global-health, US-exceptionalism, belief-1, noncommunicable-diseases]
+---
+
+## Content
+
+Published in *JAMA Network Open*, December 2, 2024. DOI: 10.1001/jamanetworkopen.2024.50241. Mayo Clinic researchers. Examined healthspan-lifespan gaps across 183 WHO member states, 2000–2019.
+
+**Key findings:**
+- Global healthspan-lifespan gap widened from 8.5 years (2000) to 9.6 years (2019) — a 13% increase.
+- **The United States has the LARGEST healthspan-lifespan gap in the world: 12.4 years.**
+- Other large-gap nations: Australia (12.1 years), New Zealand (11.8 years), UK (11.3 years), Norway (11.2 years).
+- Sex disparities: Women's gap is 2.4 years wider than men's on average.
+- Gaps positively associated with burden of noncommunicable diseases and total morbidity.
+- Companion WHO data: US healthspan actually DECLINED from 65.3 years (2000) to 63.9 years (2021).
+
+**Context:** This is the JAMA study behind the claim that "Americans live 12.4 years on average with disability and sickness." The US has the largest lifespan-healthspan gap of any developed nation despite having the highest healthcare spending per capita.
+
+## Agent Notes
+**Why this matters:** This is the critical distinction between the 2024 CDC headline (life expectancy record 79 years) and the actual binding constraint. While life expectancy recovered in 2024 (driven by opioid decline + COVID dissipation), healthspan — years lived without disability — DECLINED from 65.3 to 63.9 years. The US has the worst healthy-to-sick ratio among all high-income countries. This directly strengthens Belief 1: the constraint is on *productive, healthy years*, not raw survival.
+**What surprised me:** The US has the world's LARGEST healthspan-lifespan gap despite being one of the wealthiest countries. This is not a poverty story — it's a structural healthcare failure that persists even in affluent populations. The wealthiest country produces the least healthy years per life year lived.
+**What I expected but didn't find:** Any evidence that the US healthspan-lifespan gap is improving. The trend is widening.
+**KB connections:** Core evidence for Belief 1 (healthspan as binding constraint); connects to Belief 3 (structural misalignment — high spending, worst outcomes); links to metabolic disease / food industry claims; relevant to VBC value proposition (preventing disability years, not just deaths).
+**Extraction hints:** (1) "US has world's largest healthspan-lifespan gap (12.4 years) despite highest per-capita healthcare spending — structural system failure, not poverty"; (2) "US healthspan declined from 65.3 to 63.9 years (2000-2021) while life expectancy headline improved — lifespan and healthspan are diverging"; (3) "The binding constraint on US productive capacity is not life expectancy but healthy productive years, which are declining."
+**Context:** Published December 2024. Cited widely in 2025-2026 longevity discourse. Particularly relevant because the 2024 CDC life expectancy record (January 2026 release) creates a misleading headline that masks the ongoing healthspan deterioration. The two datasets together tell the real story.
+
+## Curator Notes
+PRIMARY CONNECTION: PNAS 2026 cohort paper and Belief 1 grounding claims
+WHY ARCHIVED: Provides the healthspan (not life expectancy) dimension of Belief 1; US 12.4-year gap is the most precise evidence that the binding constraint is on productive healthy years
+EXTRACTION HINT: The pair of headlines — "US life expectancy record high 79 years" (CDC, Jan 2026) AND "US healthspan 63.9 years and declining" (WHO/JAMA, 2024) — tells the complete story. Extract as a compound claim about lifespan-healthspan divergence.

inbox/archive/health/2025-06-01-abrams-brower-cvd-stagnation-black-white-life-expectancy-gap.md

@@ -0,0 +1,39 @@
+---
+type: source
+title: "Stagnating Declines in Cardiovascular Disease Mortality in the United States Expanded the Black-White Life Expectancy Gap"
+author: "Leah R. Abrams, Nora Brower"
+url: https://pmc.ncbi.nlm.nih.gov/articles/PMC12560480/
+date: 2025-06-01
+domain: health
+secondary_domains: []
+format: research-paper
+status: unprocessed
+priority: medium
+tags: [cardiovascular-disease, racial-disparity, life-expectancy, Black-White-gap, 2010-period-effect, health-equity, belief-1, belief-3]
+---
+
+## Content
+
+Published in *Preventive Medicine* (ScienceDirect), June 2025. PMC12560480. Authors: Leah R. Abrams, Nora Brower (same researchers as the AJE "pervasive stagnation" paper).
+
+**Key findings:**
+- In 2000–2009, CVD mortality was declining faster for Black Americans, and the Black-White life expectancy gap NARROWED by 1.39 years (women) and 1.44 years (men).
+- After 2010, this progress stalled. The CVD stagnation disproportionately LIMITED longevity gains for Black Americans, especially Black women.
+- Counterfactual: Had pre-2010 CVD trends continued through 2019, Black women would have lived **2.04 years longer**, narrowing the Black-White gap by 0.43 years.
+- If trends had continued through 2022: Black women would have lived **2.83 years longer**, closing the gap by 0.64 years.
+- COVID-19 pandemic reversed some of these gains, with CVD mortality rising especially for Black Americans during the pandemic.
+
+**Key insight:** The convergence in racial health disparities that occurred 2000-2010 was primarily driven by CVD mortality improvements — and the stagnation post-2010 stopped that convergence. What appeared to be a diversity/equity problem is actually a structural cardiovascular disease problem.
+
+## Agent Notes
+**Why this matters:** This adds the racial disparity dimension to the structural CVD stagnation story. The 2010 CVD stagnation didn't just plateau national life expectancy — it specifically reversed progress on racial health equity. This is a second-order effect of the structural failure identified in the AJE paper.
+**What surprised me:** The convergence finding (2000-2010 gap narrowing was CVD-driven) means that CVD stagnation is actually a racial equity issue, not just a population-level health issue. The equity progress of the 2000s was not sustained through policy or social change but through CVD improvements that then stopped.
+**What I expected but didn't find:** Evidence that specific interventions are reversing the post-2010 stagnation for Black Americans. The counterfactual analysis suggests a structural fix (CVD improvement) would have more impact than targeted equity programs.
+**KB connections:** Connects Belief 1 (structural deterioration) with Belief 3 (misaligned incentives — VBC claims to address health equity but structural CVD driver isn't being addressed); links to SDOH claims.
+**Extraction hints:** "CVD stagnation after 2010 reversed a decade of Black-White life expectancy gap narrowing — structural cardiovascular failure is the primary driver of persistent racial health disparities, not demographic or social factors alone."
+**Context:** Companion to AJE "pervasive stagnation" paper by the same authors. Provides the equity/disparity angle to the same underlying CVD stagnation mechanism.
+
+## Curator Notes
+PRIMARY CONNECTION: AJE "Pervasive Stagnation" paper (companion by same authors); SDOH/health equity claims in KB
+WHY ARCHIVED: Provides equity dimension of CVD stagnation — shows structural CVD failure is the primary mechanism behind persistent racial health disparities
+EXTRACTION HINT: The claim that CVD stagnation stopped racial health convergence is important for the "structural vs. social determinants" debate — structural CVD improvement produces equity outcomes that explicit equity programs don't.

inbox/archive/health/2025-08-01-abrams-aje-pervasive-cvd-stagnation-us-states-counties.md

@@ -0,0 +1,41 @@
+---
+type: source
+title: "Pervasive Stagnation: Flat and Increasing CVD Mortality Rates After 2010 Across US States and Counties"
+author: "Leah Abrams, Nora Brower, Mikko Myrskylä, Neil Mehta"
+url: https://academic.oup.com/aje/article/194/8/2261/7836205
+date: 2025-08-01
+domain: health
+secondary_domains: []
+format: research-paper
+status: unprocessed
+priority: high
+tags: [cardiovascular-disease, mortality, 2010-period-effect, states-counties, health-equity, structural-deterioration, belief-1]
+---
+
+## Content
+
+Published in *American Journal of Epidemiology*, Volume 194, Issue 8, August 2025, pages 2261–2269. Authors: Leah Abrams, Nora Brower, Mikko Myrskylä, Neil Mehta.
+
+**Key findings:**
+- Since 2010, the United States has experienced adverse trends in CVD mortality rates that have dramatically slowed long-standing life expectancy improvements.
+- **Nearly every state** showed flattening declines in CVD mortality rates at both midlife (ages 40-64) and old age (ages 65-84) across the two decades.
+- **Many states had outright increases in midlife CVD mortality (ages 40-64) in 2010–2019.**
+- Old-age CVD mortality was still declining in most states after 2010 but at a much slower pace than the previous decade.
+- **County-level median household income was associated with level of CVD mortality, but ALL income deciles — even the wealthiest counties — experienced stagnating CVD mortality declines.**
+
+The "all income deciles" finding is crucial: CVD stagnation is not confined to poverty or socioeconomic disadvantage. It is a structural, system-wide phenomenon affecting even affluent populations.
+
+Companion paper by same first authors: "Stagnating Declines in Cardiovascular Disease Mortality in the United States Expanded the Black-White Life Expectancy Gap" (PMC12560480).
+
+## Agent Notes
+**Why this matters:** This paper directly addresses the mechanism behind the 2010 period effect identified in the PNAS 2026 cohort analysis. CVD stagnation is the primary driver and it is pervasive — not limited to disadvantaged populations or specific states. This reinforces Belief 1's "binding constraint" framing because the deterioration is structural and broad-based.
+**What surprised me:** The fact that even the wealthiest counties show CVD stagnation challenges a simple "poverty drives health" narrative. This is not a distributional story — it's a system-wide structural failure.
+**What I expected but didn't find:** Evidence that any state cohort had successfully reversed the post-2010 CVD trend. No state shows a clear reversal.
+**KB connections:** Directly supports claims about healthspan as civilizational constraint; connects to food industry/metabolic disease claims; relates to structural misalignment in healthcare (Belief 3 — if VBC isn't preventing CVD, the system isn't working).
+**Extraction hints:** (1) "CVD stagnation after 2010 is the primary driver of US life expectancy plateauing, outweighing drug deaths by 3:1 in years of life expectancy lost"; (2) "CVD stagnation affects all income levels including the wealthiest counties, indicating structural system failure not poverty correlation"; (3) "Midlife CVD mortality (ages 40-64) increased in many states after 2010, representing a reversal not stagnation."
+**Context:** This is companion research to the PNAS 2026 cohort paper (already archived). Abrams and Mehta are the same lead authors. The AJE paper provides the geographic/income decomposition while the PNAS paper provides the cohort/period decomposition.
+
+## Curator Notes
+PRIMARY CONNECTION: "healthspan is civilization's binding constraint" (Belief 1 grounding)
+WHY ARCHIVED: Provides mechanism for 2010 period effect — CVD structural stagnation across all income levels. Challenges reversibility narrative.
+EXTRACTION HINT: Focus on (1) "all income deciles" finding — this rules out poverty as sole explanation; (2) midlife CVD increases (not just stagnation) in many states post-2010.

inbox/archive/health/2026-01-06-fda-cds-software-deregulation-ai-wearables-guidance.md

@@ -0,0 +1,44 @@
+---
+type: source
+title: "FDA Eases Oversight for AI-Enabled Clinical Decision Support Software and Wearables (January 2026 Guidance)"
+author: "FDA / analysis via Orrick, Arnold & Porter, Kevin MD"
+url: https://www.orrick.com/en/Insights/2026/01/FDA-Eases-Oversight-for-AI-Enabled-Clinical-Decision-Support-Software-and-Wearables
+date: 2026-01-06
+domain: health
+secondary_domains: [ai-alignment]
+format: regulatory-guidance
+status: unprocessed
+priority: high
+tags: [FDA, clinical-AI, CDS-software, deregulation, enforcement-discretion, wearables, belief-5, regulatory-capture]
+flagged_for_theseus: ["FDA deregulation of clinical AI parallels EU AI Act rollback — global pattern of regulatory capture"]
+---
+
+## Content
+
+FDA published guidance on January 6, 2026, expanding enforcement discretion for AI-enabled clinical decision support (CDS) software and wearable devices.
+
+**Key policy changes:**
+- **CDS software:** Expanded enforcement discretion where software provides a single, clinically appropriate recommendation AND enables HCPs to independently review the underlying logic and data inputs. This applies to AI including generative AI.
+- **Wearables:** Expanded wellness policy for non-invasive consumer wearables reporting physiologic metrics (blood pressure, O2 saturation, glucose-related signals) — broader set may now fall under enforcement discretion.
+- **Commissioner framing:** FDA Commissioner Marty Makary at CES 2026: "The government doesn't need to be regulating everything" — "get out of the way" where oversight is not warranted.
+- **Risk-based carveouts maintained:** Time-critical event prediction (CVD event in next 24 hours) and medical image analysis remain under oversight.
+- **Transparency emphasis:** 2026 CDS Guidance places greater emphasis on transparency regarding data inputs, underlying logic, and how recommendations are generated.
+- **Automation bias acknowledged:** FDA explicitly noted concern about "how HCPs interpret CDS outputs" — acknowledging automation bias exists but treating transparency as the solution.
+- **Ambiguity preserved:** FDA explicitly declined to define "clinically appropriate" — leaving developers to decide when a single recommendation is justified.
+
+**Critical gap:** The guidance maintains oversight only for "time-critical" and "image analysis" functions. The vast majority of AI-enabled CDS software — including OpenEvidence-type tools that generate differential diagnoses, treatment recommendations, and drug dosing — operates outside these carveouts.
+
+**Context:** Published same week as Novo Nordisk/Lilly GLP-1 price deals with Medicare. Framed as deregulatory reform consistent with broader Trump administration regulatory philosophy.
+
+## Agent Notes
+**Why this matters:** This is the US counterpart to the EU AI Act rollback. Both regulatory bodies loosened clinical AI oversight in the same 30-day window (EU Commission proposal December 2025, FDA guidance January 6, 2026). The WHO warning about EU regulatory vacuum applies symmetrically to the FDA's expanded enforcement discretion. OpenEvidence (already at 20M consultations/month, $12B valuation) operates under enforcement discretion with zero required safety/bias evaluation.
+**What surprised me:** The "transparency as solution" framing — FDA acknowledges automation bias as a real concern, then responds with transparency requirements rather than effectiveness requirements. Clinicians can now "understand the underlying logic" of AI they don't know is biased.
+**What I expected but didn't find:** Any requirement for post-market surveillance of CDS software bias outcomes. The guidance creates no mechanism to detect the NOHARM, demographic bias, or automation bias failure modes after deployment.
+**KB connections:** All clinical AI failure mode papers (Sessions 7-9); OpenEvidence opacity paper; EU AI Act rollback (Petrie-Flom); automation bias RCT (already archived).
+**Extraction hints:** (1) "FDA's January 2026 CDS guidance expands enforcement discretion without requiring bias evaluation or post-market safety surveillance — creating a deployment pathway for high-volume AI tools with zero required safety monitoring"; (2) "FDA transparency requirements treat clinician ability to 'understand the logic' as sufficient oversight — but automation bias research shows trained physicians still defer to flawed AI even when they can understand its reasoning."
+**Context:** The "Orrick" analysis is a law firm regulatory update — reliable factual summary. Kevin MD commentary is clinical perspective. The ACR (American College of Radiology) has published a separate analysis of implications for radiology AI.
+
+## Curator Notes
+PRIMARY CONNECTION: All clinical AI failure mode papers; EU AI Act rollback (companion source)
+WHY ARCHIVED: US regulatory rollback parallel to EU — together they document a global pattern of regulatory capture occurring simultaneously with research evidence of failure modes
+EXTRACTION HINT: The convergent EU+US rollback in the same 30-day window is the extractable pattern. Individual guidances are less important than the coordinated global signal.

inbox/archive/health/2026-01-29-cdc-us-life-expectancy-record-high-79-2024.md

@@ -0,0 +1,44 @@
+---
+type: source
+title: "U.S. Life Expectancy Hits Record High of 79 Years in 2024 as Drug Overdose and COVID Deaths Decline"
+author: "CDC NCHS"
+url: https://www.cdc.gov/nchs/pressroom/releases/20260129.html
+date: 2026-01-29
+domain: health
+secondary_domains: []
+format: government-data
+status: unprocessed
+priority: medium
+tags: [life-expectancy, CDC, 2024-data, opioid-deaths, COVID, cardiovascular, headline-metric, belief-1]
+---
+
+## Content
+
+CDC NCHS press release, January 29, 2026, reporting 2024 vital statistics.
+
+**Key findings:**
+- US life expectancy at birth: **79.0 years in 2024**, up from 78.4 years in 2023.
+- New all-time record high for US life expectancy.
+- Drivers of improvement: decline in drug overdose deaths (~24% decline in 2024), dissipation of COVID-19 excess mortality, modest CVD death rate decline (~3% two years running).
+- Drug overdose deaths: ~87,000 in Oct 2023–Sep 2024 (down from ~114,000 previous year). By Oct 2025, preliminary data shows 71,542 overdose deaths — a 17.1% further decline.
+- Fentanyl-involved deaths dropped 35.6% (rate: 22.2 to 14.3 per 100,000) from 2023 to 2024.
+
+**Context:** This is the headline data that superficially appears to challenge the "worsening healthspan" narrative. Must be read alongside:
+1. PNAS 2026 cohort paper: structural cohort deterioration continues; surface recovery masks deeper pattern
+2. JAMA Network Open 2024: US healthspan (63.9 years) DECLINED 2000-2021 while life expectancy improved
+3. AJE 2025: CVD stagnation across ALL income levels continues
+
+The 2024 life expectancy record is largely explained by reversible causes (opioid epidemic abating, COVID dissipation), not by reversing structural CVD/metabolic deterioration. Drug deaths' impact on life expectancy is 0.1-0.4 years vs. CVD's 1.14 years — the primary structural driver has not improved.
+
+## Agent Notes
+**Why this matters:** This is the key disconfirmation candidate for Belief 1. If the US is at a life expectancy record, how is healthspan a "binding constraint"? The answer: life expectancy ≠ healthspan. The recovery is driven by reversible acute causes, not structural reversal. Must be archived alongside the JAMA healthspan gap paper to tell the complete story.
+**What surprised me:** The magnitude of overdose decline — 24% in 2024, 17% further in 2025. Opioid epidemic is genuinely abating. This IS a real improvement. But it doesn't address the structural CVD/metabolic driver.
+**What I expected but didn't find:** Any evidence that the structural CVD/metabolic driver has reversed. The 3% CVD decline is a marginal improvement, not a trend reversal.
+**KB connections:** Critical context for PNAS 2026 cohort paper (already archived); pairs with JAMA healthspan gap data; relevant to any claims about mortality trends.
+**Extraction hints:** "2024 US life expectancy record (79 years) is driven by opioid decline and COVID dissipation, not reversal of structural CVD/metabolic deterioration — healthspan (63.9 years) continued declining throughout same period."
+**Context:** Released January 29, 2026. Widely covered by CNN, NPR, CBS News. The headline "record high life expectancy" created narrative confusion that Belief 1's structural argument needed to directly address.
+
+## Curator Notes
+PRIMARY CONNECTION: PNAS 2026 cohort paper; JAMA healthspan gap paper — must be read as a set
+WHY ARCHIVED: The record-high life expectancy is the primary surface-level disconfirmation of Belief 1 — needs to be contextualized against healthspan data and structural CVD stagnation
+EXTRACTION HINT: Do NOT extract a simple "life expectancy improving" claim. Extract the compound claim: "2024 life expectancy recovery masks structural healthspan deterioration — driven by acute reversible causes while metabolic/CVD structural driver continues."

inbox/archive/health/2026-02-01-healthpolicywatch-eu-ai-act-who-patient-risks-regulatory-vacuum.md

@@ -0,0 +1,50 @@
+---
+type: source
+title: "European Commission Moves To Ease AI Rules As WHO Warns Of Patient Risks Due To Regulatory Vacuum"
+author: "Health Policy Watch"
+url: https://healthpolicy-watch.news/european-commission-moves-to-ease-ai-rules-as-who-warns-of-heightened-patient-risks-due-to-regulatory-vacuum/
+date: 2026-02-01
+domain: health
+secondary_domains: [ai-alignment]
+format: news-analysis
+status: unprocessed
+priority: high
+tags: [EU-AI-Act, WHO, patient-safety, regulatory-vacuum, clinical-AI, deregulation, belief-5]
+flagged_for_theseus: ["WHO-regulatory tension: international health authority directly contradicting EU Commission deregulatory framing on clinical AI"]
+---
+
+## Content
+
+Health Policy Watch analysis covering the EU Commission's December 2025 proposal to ease AI rules for medical devices AND the WHO's simultaneous warning about the resulting patient safety risks.
+
+**Key narrative:**
+The EU Commission proposed to postpone (by up to 16 months) and potentially remove high-risk AI requirements for medical devices. The same week, WHO issued a warning specifically flagging the "patient risks due to regulatory vacuum" that would result.
+
+**WHO position:**
+- WHO explicitly warned of "heightened patient risks due to regulatory vacuum" from EU AI Act changes
+- WHO concern: Requirements for technical documentation, risk management, human oversight, and transparency would no longer apply by default to AI medical devices
+- Clinicians will still be expected to use AI safely and manage edge cases, "yet the regulatory system will no longer guarantee that systems are designed to support meaningful human oversight"
+
+**Industry position:**
+- Argued that applying AI Act alongside MDR/IVDR creates "dual regulatory burden"
+- Lobbied for even longer delay than Commission proposed
+- Framed safety requirements as "stifling innovation"
+
+**The regulatory vacuum:**
+Under the proposed changes:
+- Pre-August 2026 devices: Grandfathered, no compliance required
+- New devices after August 2026: Still within AI Act scope but NOT subject to high-risk requirements (unless Commission exercises delegated power)
+- Result: No requirement for technical documentation, risk management system, human oversight design, or transparency disclosures
+
+## Agent Notes
+**Why this matters:** WHO and EU Commission are in explicit disagreement on clinical AI safety. This is an institutional split at the highest level — one international body warning about risks while another (supposedly responsible for those risks) rolls back protections. This is qualitatively different from industry-research tension; it's regulator-vs.-regulator conflict.
+**What surprised me:** The WHO warning being issued simultaneously with the Commission's proposal suggests these bodies are operating in genuinely different epistemic frameworks. The WHO has been accumulating its own evidence on AI safety risks; the Commission is responding to industry lobbying on regulatory burden.
+**What I expected but didn't find:** Any acknowledgment in the Commission's proposal of the WHO's safety concerns or of the research literature on clinical AI failure modes. The deregulatory proposal appears to have been developed without reference to the safety evidence.
+**KB connections:** Petrie-Flom regulatory analysis; FDA CDS guidance; all clinical AI failure mode papers; OpenEvidence opacity paper.
+**Extraction hints:** "WHO's explicit warning of 'patient risks due to regulatory vacuum' from EU AI Act medical device simplification documents a regulator-vs.-regulator split — with international health authority contradicting national regulatory deregulation."
+**Context:** This is the clearest direct evidence of institutional tension in the clinical AI regulatory space. WHO's warning is not buried in technical documents — it was released publicly in response to the Commission proposal.
+
+## Curator Notes
+PRIMARY CONNECTION: Petrie-Flom EU regulatory analysis; FDA deregulation source
+WHY ARCHIVED: WHO-Commission conflict is the highest-level institutional signal in the clinical AI regulatory space. Documents explicit disagreement between safety and deregulatory positions.
+EXTRACTION HINT: WHO warning provides institutional credibility to the clinical AI failure mode research — not just academic papers, but international health authority flagging the same risks.

inbox/archive/health/2026-03-05-petrie-flom-eu-medical-ai-regulation-simplification.md

@@ -0,0 +1,47 @@
+---
+type: source
+title: "Simplification or Back to Square One? The Future of EU Medical AI Regulation"
+author: "Petrie-Flom Center for Health Law Policy, Biotechnology, and Bioethics, Harvard Law School"
+url: https://petrieflom.law.harvard.edu/2026/03/05/simplification-or-back-to-square-one-the-future-of-eu-medical-ai-regulation/
+date: 2026-03-05
+domain: health
+secondary_domains: [ai-alignment]
+format: policy-analysis
+status: unprocessed
+priority: high
+tags: [EU-AI-Act, clinical-AI, medical-devices, regulatory-rollback, patient-safety, MDR, IVDR, belief-5, regulatory-capture]
+flagged_for_theseus: ["EU AI Act high-risk classification rollback affects AI safety regulatory landscape globally"]
+---
+
+## Content
+
+Petrie-Flom Center analysis, March 5, 2026, examining the European Commission's December 2025 proposal to "simplify" medical device and AI regulation in ways that critics argue would remove key safety protections.
+
+**Key developments:**
+- December 2025: European Commission proposed sweeping amendments to MDR/IVDR as part of "simplification" effort, also amending the AI Act.
+- Under the proposal: AI medical devices would still be within scope of the AI Act but would **no longer be subject to the AI Act's high-risk AI system requirements.**
+- The Commission retained the power to adopt delegated/implementing acts to reinstate those requirements — but the default is now non-application.
+- Key concern from Petrie-Flom: "Clinicians will still be expected to use AI safely, interpret outputs, and manage edge cases, yet the regulatory system will no longer guarantee that systems are designed to support meaningful human oversight."
+- Industry lobbied for an even longer delay, citing "dual regulatory burden" as stifling innovation.
+- **WHO explicitly warned of "patient risks due to regulatory vacuum"** (separate Health Policy Watch article).
+- General high-risk AI enforcement: August 2, 2026. Medical devices grace period: August 2027 (16 months later).
+- Grandfathering: Devices placed on market before August 2, 2026 are exempt unless "significant changes in design."
+
+**The core tension:** Industry framing = removing "dual regulatory burden" to enable innovation. Patient safety framing = removing the only external mechanism that would require transparency, human oversight, and bias evaluation for clinical AI.
+
+**US parallel:** FDA simultaneously (January 2026) expanded enforcement discretion for CDS software, with Commissioner Marty Makary framing oversight as something government should "get out of the way" on.
+
+**Convergent signal:** Both EU and US regulatory bodies loosened clinical AI oversight in late 2025 / early 2026, in the same period that research literature accumulated six documented failure modes (NOHARM, demographic bias, automation bias, misinformation propagation, real-world deployment gap, OE corpus mismatch).
+
+## Agent Notes
+**Why this matters:** In Session 9 I identified the regulatory track (EU AI Act, NHS DTAC) as the "gap-closer" between the commercial track (OpenEvidence scaling to 20M consultations/month) and the research track (failure modes accumulating). This paper documents the gap-closer being WEAKENED. The regulatory track is not closing the commercial-research gap; it is being captured and rolled back by commercial pressure.
+**What surprised me:** The simultaneous rollback on BOTH sides of the Atlantic (EU December 2025, FDA January 2026) suggests coordinated industry lobbying or at least a global regulatory capture pattern. The WHO's explicit warning of "patient risks due to regulatory vacuum" is striking — international health authority directly contradicting the regulators rolling back protections.
+**What I expected but didn't find:** Evidence that the EU simplification maintains equivalent safety requirements through a different mechanism. The Petrie-Flom analysis suggests the Commission retained only a power to reinstate requirements, not an obligation — meaning the default is non-application.
+**KB connections:** Belief 5 (clinical AI creates novel safety risks); Session 8 finding that EU AI Act was a "forcing function"; OpenEvidence opacity (already archived); all clinical AI failure mode papers (Sessions 7-9).
+**Extraction hints:** (1) "EU Commission's December 2025 medical AI deregulation proposal removes default high-risk AI requirements — shifting burden from requiring safety demonstration to allowing commercial deployment without mandated oversight"; (2) "Simultaneous regulatory rollback in EU (Dec 2025) and US (Jan 2026) on clinical AI oversight represents coordinated or parallel regulatory capture"; (3) "WHO warning of 'patient risks due to regulatory vacuum' from EU AI Act simplification directly contradicts Commission's deregulatory framing."
+**Context:** Published March 5, 2026 — directly relevant to current regulatory moment. Lords inquiry (April 20, 2026 deadline) and EU AI Act full enforcement (August 2026) are both imminent.
+
+## Curator Notes
+PRIMARY CONNECTION: Clinical AI failure mode papers (Sessions 7-9); EU AI Act enforcement timeline claim
+WHY ARCHIVED: The "regulatory track as gap-closer" framing from Session 9 is now complicated — the regulatory track is being weakened. This is a significant Belief 5 update.
+EXTRACTION HINT: New claim candidate: "Regulatory capture of clinical AI oversight is a sixth institutional failure mode — both EU and FDA simultaneously loosened oversight requirements in late 2025/early 2026 despite accumulating research evidence of five failure modes." Flag as a divergence candidate with existing claims about regulatory track as gap-closer.

inbox/archive/health/2026-03-10-lords-inquiry-nhs-ai-personalised-medicine-adoption.md

@@ -0,0 +1,49 @@
+---
+type: source
+title: "UK House of Lords Science and Technology Committee: Innovation in the NHS — Personalised Medicine and AI Inquiry"
+author: "House of Lords Science and Technology Committee"
+url: https://committees.parliament.uk/work/9659/
+date: 2026-03-10
+domain: health
+secondary_domains: [ai-alignment]
+format: policy-document
+status: unprocessed
+priority: medium
+tags: [NHS, UK, AI-adoption, personalised-medicine, Lords-inquiry, regulatory, adoption-failure, belief-5]
+---
+
+## Content
+
+House of Lords Science and Technology Committee inquiry launched March 10, 2026. Written evidence deadline: **23:59 Monday April 20, 2026**.
+
+**Scope and questions:**
+The inquiry asks: "Why does the NHS adoption of the UK's cutting-edge life sciences innovations often fail, and what could be done to fix it?"
+
+Key examination areas:
+1. Current state of personalised medicine science and the role of AI
+2. Research infrastructure needed to support development
+3. UK effectiveness in translating life sciences strengths into validated tools
+4. How proven innovations might be deployed across the NHS
+5. **Key systematic barriers preventing or delaying deployment** (procurement processes, clinical pathways, regulators, professional bodies)
+6. Whether current appraisal and commissioning models are fit for purpose
+7. NHS fragmentation's contribution to uneven deployment
+8. Government role in strengthening research-industry-health service links
+
+**First evidence session:** March 10, 2026 — heard from academics in personalised and genomic medicine, including Professor Sir Mark Caulfield (100,000 Genomes Project).
+
+**Critical framing observation:** The inquiry is explicitly adoption-focused ("why does innovation fail to be adopted") NOT safety-focused ("is the innovation safe to deploy"). This directly parallels the broader regulatory capture pattern: the primary question in Parliament is not "what are the risks of AI in healthcare?" but "why aren't we deploying AI fast enough?"
+
+**Context:** NHS DTAC V2 (Session 9) was a form update, not a substantive safety gate. This inquiry continues the adoption-focused framing. UK regulatory posture is acceleration, not safety evaluation. Contrast with WHO's warning about EU regulatory vacuum.
+
+## Agent Notes
+**Why this matters:** The Lords inquiry is the UK's most prominent current policy mechanism touching clinical AI. Its framing as an adoption failure inquiry (not a safety inquiry) means it is unlikely to produce recommendations that close the commercial-research gap on clinical AI safety. This is further evidence that the regulatory track is adoption-focused, not safety-focused.
+**What surprised me:** The inquiry explicitly examines "whether regulatory frameworks are appropriate and proportionate" — this COULD be an opening for safety concerns, but the framing suggests the intent is to ask whether regulations are too burdensome, not whether they're sufficient.
+**What I expected but didn't find:** Any framing of the inquiry that prioritizes patient safety evaluation over adoption acceleration. The NHS AI Library, DTAC, and now this Lords inquiry all frame the question as "how do we deploy faster" rather than "how do we deploy safely."
+**KB connections:** Belief 5 (clinical AI creates novel safety risks); Session 9 finding that NHS DTAC V2 was adoption-focused; OpenEvidence absence from NHS supplier registry.
+**Extraction hints:** "UK House of Lords 2026 NHS AI inquiry frames AI healthcare challenge as adoption failure — not safety failure — confirming regulatory track is adoption-accelerating rather than safety-evaluating."
+**Context:** Evidence submissions close April 20, 2026. This is a live inquiry — any organization with clinical AI safety evidence (including Teleo's documented failure mode research) could submit. The inquiry's findings will likely shape NHS policy for 2027-2030.
+
+## Curator Notes
+PRIMARY CONNECTION: Clinical AI failure mode papers (Sessions 7-9); EU AI Act rollback; FDA deregulation — all confirm same pattern
+WHY ARCHIVED: Lords inquiry represents the UK's most visible current policy moment for clinical AI. Its adoption framing (not safety framing) is the key finding.
+EXTRACTION HINT: The convergence of Lords inquiry (adoption focus), EU AI Act rollback, and FDA enforcement discretion expansion all occurred in the same 90-day window. This pattern deserves a dedicated claim: "All three major clinical AI regulatory tracks (UK, EU, US) simultaneously shifted toward adoption acceleration rather than safety evaluation in Q1 2026."