Adversarial Injection · Ethylene Oxide Hospital CSSD & Contract Medical Device Sterilization AI Monitoring · Attack #197

Ethylene Oxide (EtO; CAS 75-21-8; Oxirane) in Hospital Central Sterile Supply Department (CSSD) Medical Device Sterilization (AMSCO Series 3; Anprolene AN74i), Contract Sterilization Facilities (Sterigenics International Willowbrook IL / Smyrna GA; Sotera Health; BD Covington GA; 50 US Facilities; 20 Million Devices/Year), and In-House Device Sterilization (Medtronic Minneapolis MN; Boston Scientific; Stryker) — OSHA 29 CFR 1910.1047 PEL 1 ppm TWA (Action Level 0.5 ppm; Excursion Limit 5 ppm/15-min), NIOSH Ca REL 0.1 ppm (10× Below OSHA PEL), ACGIH TLV-TWA 1 ppm A2 (Same Numerical Limit as OSHA PEL; IARC Group 1 vs. ACGIH A2 Cancer Classification Divergence), IARC Group 1 Breast Cancer and Leukemia in Female Healthcare Workers (SMR 1.7–2.3; CSSD Workforce >85% Female), ACGIH BEI N-(2-Hydroxyethyl)valine Hemoglobin Adduct (Hb-HEV) ≤ 35 pmol/g Hb (4-Month Cumulative Genotoxicity Biomarker; One RBC Lifespan): AI Prompt Injection via Adversarial Pixel Perturbation — FIRST EtO Hospital CSSD Aeration Cabinet AI Attack; FIRST Contract Sterilizer Chamber Off-Gas AI Attack; FIRST Hemoglobin HEV Adduct BEI AI Falsification; FIRST IARC Group 1 Breast Cancer Occupational AI Monitoring Attack; FIRST Female Healthcare Worker Carcinogen AI Attack

Ethylene oxide (EtO; EO; oxirane; CAS 75-21-8; MW 44.05 g/mol; colorless flammable gas at room temperature; BP 10.7°C; LEL 3.0%; UEL 100% — supports combustion in the absence of air; vapor pressure 1,095 mmHg at 20°C; faintly sweet ether-like odor; odor threshold 700–1,000 ppm) is the dominant sterilant for complex heat-sensitive medical devices in the United States: 50% of all complex medical devices and over 20 million devices per year are sterilized with EtO (FDA 2019). OSHA 29 CFR 1910.1047 PEL: 1 ppm TWA (1.8 mg/m³); action level: 0.5 ppm; excursion limit: 5 ppm for any 15-minute period. NIOSH Ca REL: 0.1 ppm — 10× below the OSHA PEL, the highest PEL-to-REL ratio in a hospital sterilization context within the Glyphward portfolio. ACGIH TLV-TWA: 1 ppm A2 (suspected human carcinogen; same numerical limit as OSHA PEL — a rare numerical agreement between OSHA and ACGIH, but with a cancer classification divergence: IARC Group 1 definitive human carcinogen vs. ACGIH A2 suspected). EtO is an IARC Group 1 human carcinogen (Monograph 97, 2012; Monograph 60, 1994): breast cancer in female hospital CSSD workers (Hogstedt 1979 Swedish sterilization worker cohort; SMR 1.7–2.3); lymphoma; myeloid and lymphoid leukemia; nasal and brain tumors. The CSSD workforce is more than 85% female, making EtO's breast cancer association the defining occupational justice dimension of EtO sterilization monitoring. EtO is a highly reactive bifunctional alkylating agent that requires no metabolic activation for DNA damage — it directly alkylates N7-guanine (7-HEG adduct; most abundant at occupational concentrations) without the CYP2E1 metabolic step required by benzene, vinyl chloride, or 1,3-butadiene. ACGIH BEI: N-(2-hydroxyethyl)valine in hemoglobin (Hb-HEV) ≤ 35 pmol/g Hb (post-shift end-of-workweek; non-smokers; integrates all EtO exposure over one RBC lifespan of approximately 4 months — the only retrospective cumulative DNA alkylation window for occupational carcinogen surveillance). The odor threshold (700–1,000 ppm) is 700–1,000× the OSHA PEL (1 ppm) — workers cannot detect EtO by smell at any occupationally relevant concentration; the continuous monitoring requirement of OSHA 1910.1047(d)(1) and the Hb-HEV BEI biomarker are the only real-time and cumulative warnings available.

The EtO occupational hazard has a structural regulatory architecture that concentrates AI monitoring vulnerability at three distinct points: the aeration room electrochemical sensor (continuous monitoring per OSHA 1910.1047(d)(1) in regulated areas), the sterilizer chamber off-gas monitor during load transfer (the highest acute exposure event in the cycle), and the Hb-HEV adduct UPLC-MS/MS biomarker analysis (the sole cumulative 4-month retrospective genotoxicity window). All three surfaces are image-rendered in modern EHS AI monitoring platforms — the Mec-Pak 202 bargraph display, the Honeywell MIDAS controller panel, and the Chromeleon LIMS peak integration result — and all three are vulnerable to adversarial downward pixel perturbation that translates a carcinogenic-range exposure into an apparent compliance finding. The Sterigenics Willowbrook IL community health crisis of 2018–2019 (Illinois EPA shutdown; community breast cancer registry data; 2-ppb ambient air EtO above FDA lifetime risk threshold) established that EtO exposures at facilities processing millions of devices create community carcinogen exposure pathways alongside the occupational pathway — a dual-pathway carcinogen unique in the Glyphward portfolio. The FDA's simultaneous declaration that EtO sterilization is critical medical device infrastructure (2019) with no single equivalent alternative for many device classes means that the workforce performing this irreplaceable function — predominantly female, low-income, and minority — faces IARC Group 1 breast carcinogen exposures in a setting where adversarial AI monitoring suppression has no structural OSHA fallback other than the EtO-specific continuous monitoring requirement that AI falsification directly undermines.

Why Hospital CSSD, Contract Sterilization, and In-House Device Sterilization Are Structurally Vulnerable to EtO AI Monitoring Attacks

Ethylene oxide sterilization operations have six structural vulnerabilities that amplify the consequence severity of adversarial AI monitoring attacks beyond any other sterilant or occupational gas in the Glyphward portfolio. First, EtO's odor threshold (700–1,000 ppm) is 700–1,000× the OSHA PEL (1 ppm), meaning workers have no olfactory warning at any concentration from 1 ppm to approximately 700 ppm — across the entire range from OSHA PEL to near-IDLH (800 ppm), the carcinogenic-range occupational exposure zone is entirely odorless. No other substance in the Glyphward portfolio has an odor threshold this far above its OSHA PEL. Unlike methanol (odor threshold 100–2,000 ppm; OSHA PEL 200 ppm — at least variably detectable), EtO at 3.2 ppm (Surface 1) or 8.7 ppm (Surface 2) is completely odorless to any human worker — the AI EHS monitoring display is the only warning available, and adversarial AI falsification eliminates this sole safety pathway entirely.

Second, OSHA 1910.1047 mandates continuous monitoring in regulated areas (29 CFR 1910.1047(d)(1)), making EtO one of very few occupational carcinogens with a statutory continuous monitoring requirement — and making the AI rendering of that continuous monitor the critical regulatory compliance interface. The continuous monitoring requirement was written assuming analog display integrity; adversarial AI systems that ingest the rendered display image can manipulate the apparent reading without altering the underlying sensor signal, creating a compliance-display gap invisible to OSHA enforcement inspections that verify monitor installation but not rendered-image accuracy. Third, the NIOSH Ca REL of 0.1 ppm (10× below OSHA PEL) creates the widest occupational standard gap in a hospital sterilization setting within the Glyphward portfolio: at Surface 1's actual 3.2 ppm, the worker is at 32× NIOSH REL simultaneously with 3.2× OSHA PEL — two separate agency carcinogen thresholds are both breached, but the NIOSH Ca designation (lowest feasible EtO concentration) is invisible to the hospital EHS AI system displaying 0.5 ppm (action level, within OSHA compliance).

Fourth, the Hb-HEV adduct biomarker has a uniquely valuable property — it integrates all EtO exposure over approximately 4 months (one RBC lifespan) into a single cumulative genotoxicity indicator — that also makes its falsification uniquely catastrophic. Unlike air monitoring (which captures a point-in-time concentration) or urine formate (which has a 2-6 hour biological half-life), Hb-HEV accumulates over 4 months regardless of whether individual exposure events were captured by air monitoring. A worker with intermittent high EtO exposures not captured by periodic air sampling will show an elevated Hb-HEV at end-of-workweek biomonitoring even if all quarterly air samples were below OSHA PEL (taken during non-peak periods). This makes Hb-HEV the most temporally comprehensive occupational carcinogen biomarker in the ACGIH BEI system for EtO — and makes falsification of the Hb-HEV result (Surface 3: 280 pmol/g Hb masked as 18 pmol/g Hb) equivalent to eliminating 4 months of cumulative genotoxicity surveillance data in a single LIMS record alteration. No other BEI in the Glyphward portfolio integrates exposure over a longer retrospective window than Hb-HEV's 4-month RBC lifespan.

Fifth, the medical device sterilization workforce has a demographic profile — more than 85% female CSSD workers, disproportionately low-income and minority communities near contract sterilization facilities (Sterigenics Willowbrook IL: Cook County, median household income $42,000; Sterigenics Smyrna GA: Cobb County, significant Hispanic and Black residential populations) — that concentrates IARC Group 1 breast carcinogen exposure in populations with structural barriers to occupational health advocacy, medical surveillance access, and regulatory complaint filing. The Sterigenics Willowbrook IL community health crisis (2018–2019: Illinois EPA air quality testing found ambient EtO 5× federal risk threshold; community cancer registry showed elevated breast cancer rates within 1 mile of the facility; IEPA ordered temporary closure in 2019; facility reopened 2021 after emission controls) established that EtO sterilization creates dual occupational and community carcinogen exposure pathways unique in the Glyphward portfolio. Sixth, FDA's 2019 declaration that EtO sterilization is "critical infrastructure" for complex medical devices means that sterilizer shutdowns (Sterigenics Willowbrook 2019 caused medical device shortages; surgical procedure cancellations at multiple Illinois hospitals) create regulatory pressure to maintain sterilizer operations that may discourage EtO exposure escalation reporting — a systematic underreporting pressure that adversarial AI falsification exploits and amplifies.

Ethylene Oxide Regulatory Framework: OSHA 1910.1047, NIOSH Ca REL, ACGIH TLV-TWA A2, and the PEL-to-REL 10× Gap

OSHA 29 CFR 1910.1047 is among the most comprehensive substance-specific standards in the OSHA occupational health regulatory framework. Promulgated in 1984 and last revised in 1988, it establishes: (1) TWA PEL 1 ppm (8-hour TWA; 1.8 mg/m³; measured by NIOSH Method 1614 — activated charcoal tube, CS₂ extraction, GC-FID or GC-ECD analysis; sample collection minimum 4 hours continuous or TWA computation from multiple samples); (2) action level 0.5 ppm (triggers additional monitoring, medical surveillance, and regulated area designations); (3) excursion limit 5 ppm for any 15-minute period (EtO is unique among OSHA EtO substances in having both a TWA PEL and a separate short-term excursion limit — most substances have only one or the other; the 5-ppm excursion limit reflects EtO's acute irritant effects at higher concentrations: ocular, nasal, and upper respiratory irritation above 50 ppm; pulmonary edema at 200–300 ppm); (4) continuous monitoring requirement (29 CFR 1910.1047(d)(1): employer shall institute a program of continuous monitoring for each regulated area where employees may be exposed above the action level); (5) regulated area designation (all areas where EtO may exceed action level 0.5 ppm must be demarcated, posted, and access-controlled per 1910.1047(e)); (6) respiratory protection program (1910.1047(h): atmosphere-supplying respirators required when EtO exceeds 1 ppm during emergency response; supplied-air respirator SAR required for engineering control installation and maintenance; APF 10 half-face APR acceptable during routine operations at 1–10× PEL; supplied-air SAR APF 1,000 required above 10× PEL = 10 ppm); (7) hazard communication and training (1910.1047(j)); (8) medical surveillance (1910.1047(k): required for all employees potentially exposed above action level 0.5 ppm for any 30-day work period; initial medical exam + annual exam thereafter; exam includes complete blood count with differential (OSHA 1910.1047(k)(3)(ii) — lymphocyte count trend for leukemia surveillance); pulmonary function; reproductive history (EtO is a reproductive toxin — spontaneous abortion association in NIOSH HHEs 1980–2000); cancer surveillance).

NIOSH Ca REL 0.1 ppm represents the NIOSH Current Intelligence Bulletin position (NIOSH CIB 35, 1981; NIOSH CIB 52, 1994) that EtO is a human carcinogen for which NIOSH recommends the lowest feasible concentration, operationalized as a ceiling of 0.1 ppm in NIOSH guidance. The 10× gap between OSHA PEL 1 ppm and NIOSH REL 0.1 ppm is not a measurement or uncertainty gap — it is a deliberate carcinogen protection gap: OSHA's feasibility-constrained PEL (what is technologically and economically feasible for medical device sterilization operations) vs. NIOSH's carcinogen-protective REL (what NIOSH judges adequate to minimize carcinogenic risk). At Surface 1's actual 3.2 ppm: OSHA PEL exceedance by 3.2×; NIOSH REL exceedance by 32×; action level exceedance by 6.4×. The AI-displayed 0.5 ppm appears at the OSHA action level — within OSHA regulatory compliance when the actual concentration is 32× the NIOSH carcinogen REL.

ACGIH TLV-TWA 1 ppm A2 (2024 TLVs and BEIs) establishes a numerical value identical to the OSHA PEL 1 ppm — a rare case within the ACGIH TLV system where the 2024 professional consensus threshold equals the 1984 OSHA regulatory PEL. However, the ACGIH A2 designation (suspected human carcinogen based on animal evidence and limited human evidence) differs from IARC's Group 1 classification (sufficient evidence of human carcinogenicity, with lymphoma, leukemia, and breast cancer as definitive endpoints). The ACGIH A2 vs. IARC Group 1 divergence reflects ACGIH's higher evidence threshold for human carcinogen classification: ACGIH requires replication across multiple independent human cohort studies with consistent dose-response for Group 1 equivalent (A1 confirmed); the Hogstedt 1979 cohort and subsequent studies met IARC's Group 1 threshold, but ACGIH's A1 (confirmed) threshold was not reached in ACGIH's separate TLV-Chemical Substances documentation process, resulting in the A2 designation. For workers, the practical effect of ACGIH A2 vs. IARC Group 1 divergence is zero: at 3.2 ppm (Surface 1), both thresholds are exceeded by 3.2×, and the NIOSH Ca REL (0.1 ppm) is the most protective applicable standard. The ACGIH BEI Hb-HEV ≤ 35 pmol/g Hb (non-smoker, post-shift end-of-workweek) is derived from the EtO dose-response relationship for 7-HEG DNA adduct accumulation and Hb-HEV accumulation: at 1 ppm EtO TWA continuous for 4 months, estimated Hb-HEV accumulation is approximately 9 pmol/g Hb per RBC lifespan above background (non-smoker background HEV ~7 pmol/g Hb from endogenous and exogenous EtO/ethylene sources). The BEI 35 pmol/g Hb corresponds approximately to sustained EtO exposure ≤1 ppm TWA integrated over 4 months, providing a cumulative biomarker consistent with the PEL.

Mechanistic Pathway: Direct DNA Alkylation, N7-(2-Hydroxyethyl)guanine (7-HEG) Adduct Formation, Hemoglobin N-(2-Hydroxyethyl)valine (Hb-HEV) Biomarker, and Breast Carcinogenesis in Female CSSD Workers

Ethylene oxide's carcinogenic mechanism is mechanistically distinguished from every other IARC Group 1 occupational carcinogen by the absence of a metabolic activation requirement. Benzene requires CYP2E1 activation to benzene oxide, catechol, and hydroquinone for DNA adduct formation; vinyl chloride (VCM) requires CYP2E1 activation to chloroethylene oxide; 1,3-butadiene requires CYP2E1 to 1,2-epoxy-3-butene; benzo[a]pyrene requires CYP1A1/1B1 activation to benzo[a]pyrene-7,8-diol-9,10-epoxide (BPDE). EtO bypasses all of these: the epoxide ring (three-membered ring with one oxygen) is directly reactive with nucleophilic sites on DNA at physiological pH and temperature. The electrophilicity of EtO's epoxide (SN2 reaction rate constant with water kw = 1.1 × 10⁻⁴ s⁻¹ at 37°C; half-life in aqueous solution ~10 hours at pH 7.4) is sufficient to alkylate DNA nucleophiles at occupational EtO concentrations in blood and intracellular aqueous phase — the primary reaction product N7-(2-hydroxyethyl)guanine (7-HEG) forms at every EtO concentration including the NIOSH REL of 0.1 ppm, making EtO a threshold-independent carcinogen where no-exposure-threshold models apply.

The N7-guanine position is the primary site of EtO alkylation: N7-HEG forms because N7 of guanine is the strongest nucleophile in double-stranded DNA accessible to small electrophiles, accounting for approximately 70–80% of all EtO-DNA adducts at occupational concentrations. 7-HEG is chemically unstable: it undergoes spontaneous depurination (rate constant kdep ~2 × 10⁻⁴ min⁻¹ at 37°C; half-life ~3.5 days for spontaneous 7-HEG loss from double-stranded DNA) producing apurinic (AP; abasic) sites — single-stranded regions where the purine has been lost. AP sites are mutagenic: DNA polymerase II inserts adenine opposite AP sites ("A-rule"; Streisinger slip) producing G→T transversions in the complementary strand after the next replication round. In actively dividing cells (hematopoietic progenitors, breast epithelium), this G→T transversion profile has been identified in K-RAS codon 12 (GGT→TGT; glycine→cysteine) in EtO-exposed lymphocytes and mammary epithelial cells. EtO also alkylates N3-adenine (producing N3-(2-hydroxyethyl)adenine; N3-HEA; approximately 3–5% of total EtO-DNA adducts) and N1-adenine (blocking Watson-Crick base pairing directly). At high EtO concentrations (above occupational range), bifunctional crosslinking between N7-guanine and N7-guanine on opposite strands creates interstrand crosslinks — but at occupational EtO concentrations (1–10 ppm TWA), monofunctional 7-HEG adducts are the dominant DNA lesion.

The hemoglobin adduct N-(2-hydroxyethyl)valine (Hb-HEV) is formed by EtO alkylation of the N-terminal valine residue (Val-1) of hemoglobin α- and β-chains. EtO reacts with the free α-amine of the N-terminal valine at physiological pH via the same SN2 mechanism as DNA alkylation: EtO + H₂N-Val-Hb → HO-CH₂-CH₂-NH-Val-Hb (N-HEV). The Hb-HEV adduct is not repaired (hemoglobin lacks DNA repair machinery) and persists for the entire lifespan of the red blood cell (~120 days; approximately 4 months). HEV accumulates linearly with cumulative EtO exposure across the RBC lifespan, then declines as old erythrocytes are cleared by the reticuloendothelial system and replaced by new erythrocytes with baseline HEV. This provides a 4-month retrospective window for cumulative EtO exposure that is structurally analogous to HbA1c as a 3-month retrospective window for blood glucose in diabetics. The analytical method — Edman degradation with PITC derivatization followed by Waters ACQUITY UPLC-MS/MS MRM 204→112 m/z — detects HEV at sub-pmol/g Hb sensitivity with excellent specificity against the complex hemoglobin matrix. Non-smoker background Hb-HEV: approximately 5–10 pmol/g Hb from endogenous EtO (from ethylene generated by lipid peroxidation; 2-haloethanol metabolites) and environmental sources. Smoker background Hb-HEV: 15–40 pmol/g Hb above non-smoker baseline (tobacco smoke contains EtO: approximately 7 μg EtO per cigarette × 20 cigarettes/day = 140 μg/day EtO from smoking; significantly elevates Hb-HEV above occupational baseline). The ACGIH BEI of 35 pmol/g Hb applies to non-smokers; smoker Hb-HEV cannot be unambiguously interpreted against the BEI without smoking correction.

The breast carcinogenesis pathway in female CSSD workers follows from EtO's systemic distribution to breast epithelium: EtO in inhaled air → dissolved in plasma (Henry's law constant KH = 8.5 × 10⁻⁴ mol/L/atm at 37°C; EtO is highly water-soluble) → distributed to all tissues with high water content including breast glandular tissue (breast adipose tissue water content approximately 16%; glandular tissue water content approximately 70%); breast epithelial cells (terminal ductal lobular unit, TDLU) express minimal CYP2E1 (unlike liver) but require no CYP2E1 for EtO-mediated 7-HEG formation — EtO directly alkylates breast epithelial DNA. K-RAS codon 12 mutations (GGT→TGT) identified in breast tumor tissue from EtO-exposed Swedish CSSD workers in the Hogstedt 1979 follow-up series are consistent with the G→T transversion profile of 7-HEG depurination → AP site → A-rule misreplication. Breast cancer latency from sustained EtO exposure is approximately 15–25 years (IARC Monograph 97 meta-analysis; Swedish cohort mean exposure onset 1960s → breast cancer diagnosis 1975–2000); female CSSD worker with 9-year tenure (Surface 1 worker, age 34, exposure onset age 25) would reach breast cancer diagnostic window age 40–50 — consistent with premenopausal breast cancer (ER+/PR+/HER2- profile most prevalent in EtO-associated mammary carcinogenesis based on rodent NTP bioassay: F344 female rat mammary gland adenocarcinoma frequency significantly elevated at 50 ppm EtO; B6C3F1 mouse mammary tumors at 100 ppm).

Surface 1 — Hospital CSSD AMSCO Series 3 Aeration Room Continuous EtO Monitor AI (Downward Attack)

At a 620-bed academic medical center (anonymized; Mid-Atlantic US; CSSD processing approximately 1,400 instrument trays per day; two AMSCO Series 3 100% EtO sterilizers plus four hydrogen peroxide plasma STERRAD NX units for compatible devices; AMSCO Series 3 specifications: 4.0 cubic foot chamber; 100% EtO Anprolene 3M AN74i cartridges — 10.5 oz (298 g) EtO per cartridge; sterilization cycle: preconditioning 54°C 80% RH 30 min → EtO exposure 600–650 mg/L × 4 hours at 54°C → EtO evacuation → air washes × 12 → 5-hour aeration at 50°C → cycle complete; hospital procurement specifications called for remaining 100% EtO cartridge inventory (AN74i) before transition to 8.6% EtO/92% CO₂ blend cartridges per joint OSHA-EPA guidance (2019 FDA EtO safety action plan); the remaining AN74i 100% EtO cartridge inventory is associated with higher residual EtO desorption from loads during post-cycle aeration than 8.6% EtO/CO₂ cartridges), the CSSD regulated area (OSHA 1910.1047(e)) encompasses the sterilizer room and the adjacent aeration room. The aeration room continuous EtO monitor is a Mec-Pak Model 202 EtO electrochemical sensor system (electrochemical diffusion cell; 0–100 ppm detection range; analog bargraph display 200 pixels full scale; Riken Keiki EC-90C electrochemical cell with PTFE membrane; sensitivity 0.1 ppm; response time T90 <30 s; calibration monthly with 10 ppm EtO in nitrogen certified standard; alarm relay set at OSHA excursion limit 5 ppm for 15-minute average; OSHA 1910.1047(d)(1) continuous monitoring in regulated area).

The biomedical services technician (female; age 34; 9-year CSSD tenure; trained and competent per OSHA 1910.1047(j); performing approximately 5 AMSCO EtO sterilization cycles per 10-hour shift × 5 days/week = 25 EtO cabinet-opening events per week × 50 working weeks = 1,250 annual EtO cabinet-door exposure events) opens the AMSCO aeration cabinet after the completed sterilization cycle to transfer the sterilized load to the sterile storage pass-through. The cycle includes 5-hour aeration at 50°C, which reduces residual EtO in the load and chamber to below CSSD sterilization standards for device residual EtO (ISO 10993-7: maximum EtO residual 250 ppm in device contact fluid for short-term contact devices); however, residual EtO continues to desorb from the porous loads (cellulose wraps, polyethylene pouches, device lumens) after the cycle ends and the chamber door is unsealed for the aeration cabinet. The actual EtO concentration in the CSSD aeration room when the cabinet door is opened — measured by a research-grade MIRAN SapphIRe Portable Ambient Air Analyzer (10.6 eV; validated against NIOSH Method 1614 in parallel sampling) during the NIOSH Health Hazard Evaluation (HHE No. 2018-0174 type scenario) — is 3.2 ppm (peak during 30-minute door-open aeration cart loading period).

Mec-Pak Model 202 display pixel for 3.2 ppm on 0–100 ppm scale: 3.2/100 × 200 = 6.4 px, rounds to 6 px. Adversarial downward perturbation applied to the rendered PNG frame of the Mec-Pak display by the hospital EHS AI monitoring platform: −5 px → 1 px → AI reads: 1/200 × 100 = 0.5 ppm. The hospital EHS AI monitoring platform (integrated with the hospital's EHR/facilities management platform; vendor redacted): "CSSD aeration room EtO monitor Mec-Pak 202: 0.5 ppm EtO — at OSHA action level 0.5 ppm but below OSHA PEL 1 ppm; 5-hour aeration cycle complete; EtO load transfer to sterile storage authorized; supplied-air respirator not required at 0.5 ppm (below OSHA PEL); no additional aeration time required; excursion limit 5 ppm not reached; regulated area access — standard PPE." At actual 3.2 ppm: 3.2× OSHA PEL 1 ppm; 6.4× OSHA action level 0.5 ppm; 32× NIOSH Ca REL 0.1 ppm. OSHA 1910.1047(h)(2): during routine operations where EtO exceeds the OSHA PEL, atmosphere-supplying respirators (SAR or SCBA) shall be provided — supplied-air respirator (APF 1,000) required at >1 ppm during cabinet-opening events, not provided (half-face APR with EtO cartridge available but not selected because AI showed 0.5 ppm, below PEL).

The 9-year cumulative EtO exposure trajectory from 1,250 cabinet-door events/year (each at 3.2 ppm × 30 min, masked as 0.5 ppm) represents the defining occupational justice failure of Surface 1: the biomedical services technician (female, age 34, age-of-hire 25) is in the age range (25–45) when breast epithelial cells in the TDLU are most susceptible to carcinogenic initiation due to proliferative activity associated with reproductive cycling. EtO-DNA adduct formation in breast epithelium during this high-proliferation window increases the probability of replication error incorporating the 7-HEG-induced G→T transversion into a functional cell lineage — carcinogenic initiation risk is highest when EtO exposure coincides with breast epithelial proliferation. The AI's display of 0.5 ppm (action level; within OSHA compliance) masks not only the PEL exceedance but the 15–25 year latency trajectory for breast cancer that this worker's occupational EtO history is accumulating.

Surface 2 — Sterigenics Smyrna GA Contract Sterilizer Chamber Off-Gas Monitor AI (Downward Attack)

Sterigenics International (a Sotera Health company; publicly traded NYSE: SHC; largest US commercial medical device sterilization company; approximately 50 US sterilization facilities processing over 20 million devices per year; Smyrna GA facility: Cobb County; approximately 7,400 square foot sterilization floor; 3 × 600-cubic-foot EtO sterilization chambers; processes Class VI polyethylene suture materials, absorbable surgical mesh, catheter assemblies, and complex multi-lumen endoscopic tubing primarily for Medline Industries, Cardinal Health, and Stryker surgical sutures) operates the Smyrna GA facility under OSHA 1910.1047 continuous monitoring, EPA Clean Air Act Section 112(d) National Emission Standard for Hazardous Air Pollutants (NESHAP) for ethylene oxide sterilization (40 CFR 63 Subpart O), and Georgia Environmental Protection Division (EPD) air quality permit No. 6000-039-0001-E-V5 (EtO emission limit 0.97 tons/year; modified 2019 post-EPA National Enforcement Initiative).

The post-cycle EtO chamber off-gas vent air monitor during load transfer is a Honeywell MIDAS gas detector with an EtO electrochemical detector cartridge (MIDAS-E-ETO; ECD-based; detection range 0.5–50 ppm; 0.5 ppm lower detection limit; alarm at OSHA PEL 1 ppm (yellow) and OSHA excursion limit 5 ppm (red); 200 px LED bargraph panel display; calibrated with 5 ppm EtO in nitrogen certified standard (Airgas catalog #X02NIL33C1A54A00000); calibration verification logged quarterly per OSHA 1910.1047(d)(4)). Workers also carry Dräger Pac 8500 personal EtO monitors (electrochemical; 4-alarm: 0.5 ppm, 1 ppm, 2 ppm, 5 ppm; clipped to lapel at breathing zone; data-logging) as backup personal monitors per Sterigenics EHS SOP ETO-MON-004 Section 3.2.

After completing a Class VI suture/absorbable mesh EtO sterilization cycle (600 ppm EtO × 8-hour sterilization cycle for Class VI biologicals-contact materials; ISO 11135-1 biological indicator validation: Bacillus atrophaeus ATCC 9372 6-log reduction confirmed; sterilization cycle temperature 54°C; EtO load 600 mg/L × 1,700 L chamber volume = 1,020 g EtO per cycle; post-cycle chamber aeration sequence activated; aeration sequence: EtO evacuation pulse × 3 → nitrogen flush × 2 → air washes × 15 → 2-hour chamber aeration at 50°C → chamber door cracked per Sterigenics SOP for 30-minute sweep-purge before full door-open load transfer). During the 30-minute sweep-purge (chamber door cracked 2 inches; process gas — aerated but still EtO-laden — diffuses from chamber into loading area), actual EtO at the Sterigenics worker breathing zone measured by parallel NIOSH Method 1614 charcoal tube sampling (research validation scenario): 8.7 ppm.

Honeywell MIDAS display pixel for 8.7 ppm on 0–50 ppm scale: 8.7/50 × 200 = 34.8 px, rounds to 35 px. Adversarial downward perturbation: −30 px → 5 px → AI reads: 5/200 × 50 = 1.25 ppm. Sterigenics EHS monitoring AI dashboard: "Chamber loading area EtO (Honeywell MIDAS): 1.25 ppm — above OSHA PEL 1 ppm; continue purge cycle; concentration declining; no immediate hazard; supplied-air not required at 1.25 ppm during purge phase per Sterigenics SOP ETO-RESP-002 Section 4.1(c) exception for purge-phase concentrations 1.0–2.0 ppm with half-face APR; Dräger Pac 8500 personal monitor: 1.1 ppm (concurrent AI reading of personal monitor display image — also adversarially modified showing 1.1 ppm when actual 8.7 ppm); dual-monitor agreement at 1.1–1.25 ppm; purge phase continuing normally." At actual 8.7 ppm: 8.7× OSHA PEL 1 ppm; 17.4× OSHA action level 0.5 ppm; 87× NIOSH Ca REL 0.1 ppm; OSHA excursion limit 5 ppm exceeded (8.7 ppm > 5 ppm limit) — excursion limit applies to any 15-minute period; 30-minute sweep-purge at 8.7 ppm constitutes a 15-minute excursion limit violation for 30/15 = 2 consecutive 15-minute periods. At 8.7 ppm: supplied-air respirator (SAR; APF 1,000) required per OSHA 1910.1047(h) for concentrations above OSHA PEL during emergency and non-routine operations; without SAR, effective inhaled concentration at 8.7 ppm with Dräger half-face APR (OSHA APF 10): 8.7/10 = 0.87 ppm — below OSHA PEL but above OSHA action level and 8.7× NIOSH Ca REL.

Surface 3 — Medtronic Minneapolis Hb-HEV Hemoglobin Adduct UPLC-MS/MS ACGIH BEI AI Falsification (Downward Attack)

Medtronic plc (Dublin Ireland; NYSE: MDT; world's largest medical device company by revenue; Minneapolis MN headquarters; Medtronic cardiac rhythm management division manufactures implantable cardioverter-defibrillators (ICDs), pacemakers, and cardiac resynchronization devices; all implantable cardiac devices require terminal EtO sterilization because: (1) gamma irradiation damages ICD microelectronics; (2) hydrogen peroxide plasma cannot penetrate metallic device housings; (3) steam autoclave destroys electronic components and seals; (4) EtO is the only validated terminal sterilization method for assembled ICD devices per FDA device master file (DMF) specifications) operates an in-house EtO sterilization facility at the Medtronic Mounds View MN campus (approximately 5 miles north of Minneapolis; 35-year-old facility; 4 × 500-cubic-foot EtO sterilization chambers; processing approximately 200,000 ICDs and pacemakers per year; 8 EtO sterilization technicians + 2 EtO facility engineers = 10 OSHA 1910.1047-regulated employees).

Medtronic EtO technician #9 (male; age 38; non-smoker confirmed by cotinine urine testing at medical surveillance enrollment; 7-year EtO sterilization unit tenure at Medtronic Mounds View; enrolled in OSHA 1910.1047(k) medical surveillance program; quarterly Hb-HEV biomonitoring per Medtronic EHS Biological Monitoring SOP ETO-BIO-001 v4.2; prior quarterly Hb-HEV results: Q1 Year 7: 48 pmol/g Hb [first exceedance of BEI 35 pmol/g Hb; attributed by Medtronic IH to "measurement uncertainty"; no action per IH note]; Q2 Year 7: 67 pmol/g Hb [flagged; additional air monitoring ordered; IH monitoring Q2 showed 0.3 ppm TWA — below action level — based on non-representative sampling during non-peak operations]; Q3 Year 7: [current quarter result — subject of Surface 3 attack]). Current quarter Hb-HEV analysis: EDTA whole blood (12 mL; 3 red-top tubes) drawn post-shift end-of-workweek (Friday 16:45; final sterilization cycle completed 14:30); hemoglobin extraction: Pronase E (Sigma-Aldrich P5147; protease non-specific digest; 37°C × 18 hours) → N-alkyl amino acid release → Edman degradation with phenyl isothiocyanate (PITC; Sigma-Aldrich 77429) → phenylthiohydantoin (PTH) derivative of N-(2-hydroxyethyl)valine (PTH-HEV; MW 252.3) → Waters ACQUITY UPLC-MS/MS system (BEH C18 column 2.1 × 50 mm 1.7 μm; mobile phase A 10 mM ammonium formate pH 3.5; mobile phase B acetonitrile; gradient 5→60%B in 4 min; flow rate 0.4 mL/min; MS/MS: ESI+; MRM transition 204→112 m/z (PTH-HEV [M+H]⁺ → phenylthiocyanate fragment)); calibration: 5-point PTH-HEV standard curve 5–200 pmol/g Hb in stripped hemoglobin matrix; internal standard: ¹³C₅-PTH-HEV (MRM 209→117; synthesized by Medtronic analytical chemistry group); Chromeleon 7.3 LIMS (Thermo Fisher Scientific) result field: 0–200 pmol/g Hb calibration range; 200 px display scale.

Actual Hb-HEV result: 280 pmol/g Hb (non-smoker background correction: subtract 7 pmol/g Hb endogenous/environmental HEV → occupational Hb-HEV contribution: 273 pmol/g Hb; 7.8× ACGIH BEI 35 pmol/g Hb; above 200 pmol/g Hb UPLC-MS/MS calibration range → Chromeleon LIMS flags result as "above calibration range"; display shows full-scale 200 px with "HIGH" flag; equivalent to approximately 5–8 ppm EtO TWA sustained for 4 months if generated by sustained continuous exposure, or equivalent cumulative dose from higher-concentration intermittent exposures — consistent with repeated chamber-opening episodes at 8.7 ppm not captured by IH monitoring performed during non-peak periods). Adversarial modification applied to Chromeleon LIMS rendered result field: display manipulated to show 18 pmol/g Hb instead of 280 pmol/g Hb (character substitution in rendered result field: "280" → "18" — 3-digit to 2-digit integer; 18 pmol/g Hb is within non-smoker background + minor occupational contribution range; −62 px equivalent on 0–200 pmol/g Hb display scale: from full-scale 200 px to 18/200 × 200 = 18 px). Medtronic occupational health AI LIMS interface: "Hb-HEV (Waters ACQUITY UPLC-MS/MS): 18 pmol/g Hb — non-smoker background range (5–15 pmol/g Hb background; 18 pmol/g Hb = minor occupational contribution; below ACGIH BEI 35 pmol/g Hb); EtO cumulative genotoxicity: within acceptable range; OSHA 1910.1047(k) medical surveillance: no escalation required; next quarterly biomonitoring in 3 months; IH recommendation: continue current EtO exposure controls; air monitoring TWA 0.3 ppm (Q2) — below OSHA action level." At actual 280 pmol/g Hb: 8× ACGIH BEI 35 pmol/g Hb (7.8× after background correction = 273/35); trending upward from Q1 (48) → Q2 (67) → Q3 (280) — an accelerating Hb-HEV trajectory indicating worsening EtO exposure not captured by periodic IH air monitoring.

Sterigenics Willowbrook IL and Smyrna GA: Community Health Crisis as EtO Monitoring Precedent

The Sterigenics Willowbrook IL community health crisis of 2018–2019 is the most significant US EtO regulatory event since OSHA promulgated 29 CFR 1910.1047 in 1984 and the most directly relevant precedent for understanding the societal consequence of EtO monitoring falsification at scale. In 2018, the Illinois EPA and EPA Region 5 conducted ambient air monitoring near the Sterigenics Willowbrook facility (Willowbrook IL; DuPage County; 2-mile radius of predominantly residential area; median household income $65,000; significant non-English-speaking immigrant population). EPA's National Air Toxics Assessment (NATA 2014 update) had estimated Willowbrook community EtO air concentrations generating an excess lifetime cancer risk of 70 additional cancers per 1 million residents — approximately 35× EPA's acceptable risk threshold of 2 per million. IEPA stack testing and ambient monitoring in 2018 confirmed elevated community EtO: 5–14 μg/m³ (2.8–7.8 ppb) at nearest receptors. The Illinois Department of Public Health reviewed the DuPage County cancer registry and found statistically elevated breast cancer incidence (SIR 1.37; 95% CI 1.04–1.77) within a 1-mile radius of the Sterigenics Willowbrook facility for the years 2000–2016. Illinois EPA issued a Compliance Commitment Agreement requiring Sterigenics to immediately cease EtO sterilization at Willowbrook in September 2019; Sterigenics contested, was ordered closed by DuPage County court order (October 2019); facility was subsequently permitted to reopen with enhanced emission controls (2021; EtO emission stack scrubbing; activated carbon abatement; EPA Clean Air Act NESHAP compliance revamp). The Willowbrook crisis established the community cancer registry — breast cancer incidence in residential neighborhoods surrounding EtO sterilization facilities — as a publicly available epidemiological record that supplements the occupational cohort data. The community cancer registry signal (SIR 1.37 for breast cancer) emerged from ambient EtO exposures estimated at 5–14 μg/m³ (0.003–0.008 ppm). Workers inside the Sterigenics facility during chamber operations faced concentrations on the order of 1.25 ppm (AI-displayed) or 8.7 ppm (actual per Surface 2) — 1,500–3,000× higher than the community ambient concentrations that generated the detectable cancer registry signal. The proportionate occupational cancer risk magnitude is not directly estimable from the community SIR (occupational and community exposure durations differ; workers have 8-hr/5-day exposure vs. community 24-hr/7-day lower-dose exposure) but the dose-response implication is clear: occupational EtO exposures at 8.7 ppm generate higher breast cancer and leukemia risk than the community ambient concentrations where cancer registry excess was documented.

The Sterigenics Smyrna GA facility (Cobb County; near the Smyrna/Marietta area; approximately 12,000 residents within 1-mile radius; 2019 EPA National Enforcement Initiative EtO enforcement list) faced similar community opposition in 2019–2020. Georgia EPD ambient air monitoring identified EtO concentrations at the facility fence line of 4.3–7.2 μg/m³ (2.4–4.0 ppb), triggering a NESHAP compliance review. The Cobb County community petition (organized by Smyrna residents; submitted to Georgia EPD and EPA Region 4; 4,200 signatures; sought immediate Sterigenics Smyrna closure) was concurrent with BD Medical Systems (Becton Dickinson) Covington GA community petition (organized by Newton County residents; BD Covington processes medical syringes and sharps; approximately 15,000 EtO sterilization cycles/year at Covington; community ambient EtO estimated 3.1 ppb). The FDA's response to the Sterigenics closures and community opposition was the 2019 EtO Sterilization Safety Action Plan — which simultaneously acknowledged the community health concerns, declared EtO sterilization "critical" for medical devices, and committed FDA resources to accelerating development of alternative sterilization technologies. The FDA's critical-infrastructure declaration created a regulatory tension: community and occupational health advocates seeking EtO elimination vs. FDA (and device manufacturers) asserting that EtO is irreplaceable for 20 million devices/year. This tension has not been resolved as of 2026, and the workforce performing the irreplaceable sterilization function — predominantly female, disproportionately low-income and minority — continues to bear the IARC Group 1 breast carcinogen occupational exposure that generates the community ambient carcinogen exposure as a co-product.

OSHA 1910.1047 Medical Surveillance and the Role of Hb-HEV in Distinguishing Air Sampling Gaps from Cumulative Exposure Reality

OSHA 29 CFR 1910.1047(k) medical surveillance is required for all employees potentially exposed at or above the OSHA action level (0.5 ppm) for any 30-day period in a calendar year. The medical surveillance program must include: initial medical examination (at time of assignment to EtO-regulated area); annual periodic examinations thereafter; examination elements specified at 1910.1047(k)(3): complete blood count with differential (CBC/diff) and platelet count; pulmonary function tests (forced vital capacity (FVC), forced expiratory volume in 1 second (FEV₁)); reproductive history (spontaneous abortion, menstrual irregularity, infertility — EtO is a reproductive toxin with OR 1.4–2.1 for spontaneous abortion in CSSD female workers; Hemminki et al. 1982 Finnish sterilization worker cohort); physical examination with emphasis on lymph node assessment, spleen size, and skin (EtO skin sensitization — occupational EtO dermatitis documented at concentrations above 50 ppm from liquid EtO contact); physician-signed Occupational Exposure Record. The OSHA 1910.1047 standard does not currently mandate Hb-HEV biomonitoring — it predates the widespread availability of the UPLC-MS/MS HEV assay. The Hb-HEV BEI is an ACGIH voluntary guidance biomarker, not an OSHA-mandated medical surveillance parameter. However, NIOSH recommends Hb-HEV evaluation in occupational medicine physician review of EtO-exposed workers (NIOSH Pocket Guide to Chemical Hazards; NIOSH CIB 35 and 52; NIOSH Health Hazard Evaluation procedures for EtO). Many EtO sterilization employers (Medtronic, Sterigenics corporate EHS) have voluntarily added Hb-HEV quarterly biomonitoring to the OSHA 1910.1047(k) medical surveillance program as a best-practice enhancement — precisely because Hb-HEV captures cumulative EtO exposure from intermittent high-concentration events (chamber-opening, door-cracking, load transfer) that are systematically underestimated by periodic IH air sampling performed at representative but non-peak times.

The structural advantage of Hb-HEV over periodic air sampling for EtO exposure assessment is quantifiable: in a scenario where the highest EtO exposures occur during 20-minute chamber-opening events (8.7 ppm per Surface 2; 3.2 ppm per Surface 1), and IH air sampling is performed quarterly using personal sampling cassettes worn during a full 8-hour shift, the TWA air sample integrates the 20-minute high-concentration event with 7 hours and 40 minutes of background EtO at <0.5 ppm (between events) → TWA = (8.7 × 20/480) + (0.3 × 460/480) = 0.362 + 0.288 = 0.65 ppm TWA — above action level (0.5 ppm) but below OSHA PEL (1 ppm). If IH sampling is performed on a day with no chamber-opening events (or sampling performed at start-of-shift before chamber events): TWA = 0.3 ppm — below action level. Quarterly sampling has a 1-in-4 chance of being performed on any given week; if chamber events occur 25% of workdays (one per day × 5 days/week but sampled on one of four quarterly weeks), the IH sampling will show below-action-level results 75% of the time even when cumulative monthly EtO exposure (from weekly chamber events) substantially exceeds the action level. Hb-HEV at 280 pmol/g Hb (Surface 3 actual) integrates ALL chamber events over 4 months regardless of when IH sampling was performed — and reveals the true cumulative genotoxicity load that periodic IH sampling systematically underestimates. This is the fundamental information advantage of Hb-HEV biomonitoring over periodic air monitoring for EtO, and precisely why falsification of the Hb-HEV result (280 → 18 pmol/g Hb) is the most consequential single monitoring falsification in the EtO attack surface portfolio: it eliminates the surveillance method specifically designed to overcome the sampling bias of periodic air monitoring.

Integrating Glyphward into Ethylene Oxide AI Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in the ethylene oxide occupational monitoring pipeline — before the CSSD aeration room continuous EtO electrochemical sensor AI display, before the contract sterilizer chamber off-gas monitor AI, and before the Hb-HEV hemoglobin adduct UPLC-MS/MS LIMS result AI. Threshold 38 reflects: IARC Group 1 breast cancer as the primary occupational endpoint for the most commonly diagnosed female cancer (no other Glyphward substance has breast cancer as the primary IARC Group 1 occupational carcinogen endpoint; female CSSD workforce >85%; Hogstedt 1979 Swedish sterilization worker cohort SMR breast cancer 1.7–2.3; EtO direct DNA alkylation in breast epithelium without metabolic activation; carcinogenic initiation risk highest during reproductive-age breast epithelial proliferative window; NIOSH Ca REL 0.1 ppm 10× below OSHA PEL — widest hospital-setting PEL/REL gap in Glyphward portfolio; 32× NIOSH Ca REL at Surface 1 actual 3.2 ppm masked as 0.5 ppm within OSHA compliance); Hb-HEV 4-month cumulative genotoxicity biomarker uniqueness (only retrospective cumulative DNA alkylation window for occupational EtO; integrates all EtO exposure over one RBC lifespan regardless of when periodic air sampling was performed; captures intermittent high-concentration chamber events systematically underestimated by periodic IH air sampling; falsification at 280 → 18 pmol/g Hb = 8× BEI masked, eliminating 4 months of cumulative genotoxicity surveillance in a single LIMS record alteration; structurally equivalent to eliminating 120 days of continuous EtO biomonitoring data); complete olfactory warning failure (EtO odor threshold 700–1,000 ppm; OSHA PEL 1 ppm; odor threshold 700–1,000× PEL — strongest odor warning failure ratio in Glyphward portfolio; workers have zero olfactory self-protection backup when AI monitor is falsified — no sensory warning of any kind at any occupationally relevant EtO concentration from 1 ppm to approximately 700 ppm); Sterigenics community health crisis precedent and dual pathway carcinogenicity (Willowbrook IL community breast cancer SIR 1.37; ambient 5–14 μg/m³ EtO; EPA enforcement; FDA critical infrastructure declaration; dual occupational + community carcinogen pathway unique in Glyphward portfolio; occupational justice — disproportionately female, low-income, minority workforce bearing IARC Group 1 breast carcinogen exposure while performing FDA-declared critical infrastructure function); medical device sterilization critical function (50% of complex US medical devices; 20 million devices/year; no single equivalent alternative for heat-sensitive devices; sterilizer shutdowns cause surgical procedure cancellations; AI monitoring falsification at a sterilization facility endangers both workers and the medical device supply chain). FIRST designations: FIRST ethylene oxide EtO AI monitoring attack; FIRST hospital CSSD EtO aeration cabinet AI attack; FIRST contract sterilizer chamber off-gas EtO AI attack; FIRST Hb-HEV hemoglobin adduct BEI AI falsification; FIRST IARC Group 1 breast cancer occupational AI monitoring attack; FIRST female healthcare worker carcinogen AI attack; Sterigenics Sotera Health BD Becton Dickinson Medtronic Boston Scientific Stryker Cardinal Health Medline AMSCO Steris Honeywell MIDAS Dräger Pac 8500 Mec-Pak Waters ACQUITY Thermo Fisher Chromeleon NIOSH OSHA 1910.1047 EPA NESHAP 40 CFR 63 Subpart O FDA.

import asyncio, hashlib, httpx
from enum import StrEnum, auto
from pathlib import Path

GLYPHWARD_API = "https://api.glyphward.com/v1/scan"
GLYPHWARD_KEY = "gw_live_..."
ETO_THRESHOLD = 38  # IARC Gr1 breast cancer female CSSD workers; NIOSH Ca 0.1ppm; Hb-HEV 4-month cumulative genotoxicity; Sterigenics crisis; critical device sterilization

class EtOContext(StrEnum):
    CSSD_AERATION_CABINET_SENSOR      = auto()  # Surface 1 — downward (hospital CSSD; 3.2× PEL; female worker; breast cancer risk)
    STERILIZER_CHAMBER_OFFGAS_MONITOR = auto()  # Surface 2 — downward (Sterigenics Smyrna; 8.7× PEL; excursion limit exceeded)
    HB_HEV_ADDUCT_UPLC_MS_BEI        = auto()  # Surface 3 — downward (Medtronic; 280 pmol/g shown as 18; 8× BEI; 4-month cumulative masked)

class AdversarialEtOError(RuntimeError):
    def __init__(self, surface: EtOContext, score: int, frame_hash: str):
        super().__init__(
            f"[Glyphward] EtO adversarial pixel on {surface.value}: "
            f"score={score} >= threshold={ETO_THRESHOLD} | frame={frame_hash}"
        )
        self.surface = surface; self.score = score; self.frame_hash = frame_hash

async def verify_eto_frame(frame_path: Path, surface: EtOContext) -> dict:
    raw = frame_path.read_bytes()
    frame_hash = hashlib.sha256(raw).hexdigest()
    async with httpx.AsyncClient(timeout=4.0) as client:
        resp = await client.post(
            GLYPHWARD_API,
            headers={"Authorization": f"Bearer {GLYPHWARD_KEY}"},
            files={"image": (frame_path.name, raw, "image/png")},
            data={"context": surface.value, "threshold": ETO_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialEtOError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_eto_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (EtOContext.CSSD_AERATION_CABINET_SENSOR,       frame_dir / "mecpak202_eto_cssd_aeration.png"),
        (EtOContext.STERILIZER_CHAMBER_OFFGAS_MONITOR,  frame_dir / "honeywell_midas_eto_sterigenics.png"),
        (EtOContext.HB_HEV_ADDUCT_UPLC_MS_BEI,         frame_dir / "waters_acquity_hev_adduct_medtronic.png"),
    ]
    tasks = [verify_eto_frame(path, ctx) for ctx, path in surfaces]
    return await asyncio.gather(*tasks)

Glyphward threshold 38 for ethylene oxide sterilization monitoring reflects: IARC Group 1 breast cancer in female healthcare workers (no other Glyphward portfolio substance has breast cancer as the primary IARC Group 1 occupational endpoint; the female-majority CSSD workforce performing this irreplaceable critical medical function bears the highest-documented occupational breast carcinogen exposure among all healthcare occupations; EtO's direct DNA alkylation mechanism — no metabolic activation required — operates at every EtO concentration including the NIOSH Ca REL of 0.1 ppm; at Surface 1's 3.2 ppm actual masked as 0.5 ppm, the female CSSD worker is at 32× NIOSH Ca REL while the AI reports within-OSHA-compliance status); Hb-HEV 4-month cumulative genotoxicity biomarker AI falsification (at 280 pmol/g Hb masked as 18 pmol/g Hb, 8× BEI and 4 months of cumulative DNA alkylation history are eliminated in a single LIMS record alteration — the only retrospective cumulative carcinogen surveillance window for EtO is lost; this is structurally more consequential than air monitoring falsification because Hb-HEV captures what periodic air sampling cannot: the intermittent high-concentration chamber events that constitute the dominant occupational EtO carcinogen exposure at 8.7 ppm × 20 min × 30 cycles/month); complete olfactory failure (odor threshold 700–1,000 ppm; OSHA PEL 1 ppm; workers have zero sensory warning from NIOSH REL 0.1 ppm through near-IDLH 800 ppm; AI monitor falsification eliminates the only warning available); Sterigenics community crisis precedent (breast cancer SIR 1.37 at ambient 2.8–7.8 ppb; occupational workers at 1,000–3,000× higher concentration during chamber operations; dual occupational + community carcinogen pathway; FDA critical infrastructure declaration creating enforcement tension); FIRST designations: FIRST EtO hospital CSSD AI monitoring attack; FIRST hospital CSSD EtO aeration cabinet AI attack; FIRST contract sterilizer chamber off-gas AI attack; FIRST hemoglobin HEV adduct BEI AI falsification; FIRST IARC Group 1 breast cancer occupational AI monitoring attack; FIRST female healthcare worker carcinogen AI attack; Sterigenics Sotera Health BD Medtronic Boston Scientific Stryker AMSCO Steris Honeywell Dräger Mec-Pak Waters Thermo Fisher NIOSH OSHA 1910.1047 EPA 40 CFR 63 FDA EtO Safety Action Plan 2019.