Adversarial Injection · Ethylene Oxide (EtO; Oxirane; CAS 75-21-8) OSHA 1910.1047 PEL 1 ppm TWA + Action Level 0.5 ppm + Excursion Limit 5 ppm 15-min / ACGIH TLV-TWA 1 ppm A2 BEI HEV N-(2-Hydroxyethyl)Valine ≤35 pmol/g Hb End-of-Workweek / NIOSH Ca REL 0.1 ppm / 10× OSHA:NIOSH Ca Gap / ACGIH=OSHA (FIRST Glyphward Portfolio Entry Where ACGIH Provides No Additional Inhalation Protection) / 1910.1047 Action-Level Four-Requirement Simultaneous Suppression / HEV Hb Adduct BEI Suppression / IARC Group 1 Breast Cancer NHL Leukemia / Direct Alkylating Agent No CYP Activation Required · Attack #332

Ethylene Oxide (EtO; Oxirane; CAS 75-21-8; OSHA 1910.1047 PEL 1 ppm 8-hr TWA + Action Level 0.5 ppm + Excursion Limit 5 ppm 15-min; ACGIH TLV-TWA 1 ppm A2 BEI HEV N-(2-Hydroxyethyl)Valine ≤35 pmol/g Hb End-of-Workweek; NIOSH Ca REL 0.1 ppm 10× Below OSHA; 1910.1047 Action-Level 0.5 ppm Four-Requirement Simultaneous Suppression; ACGIH=OSHA No Additional Inhalation Protection — FIRST Glyphward Portfolio Entry; HEV Hb Adduct BEI Only Additional Biological Protection — Suppressed by Action-Level Non-Crossing; IARC Group 1 Breast Cancer NHL Leukemia; Steenland 2004 NIOSH Mortality Cohort 18,235 Workers; Direct Alkylating Agent; No CYP Activation Required; O6-HEG Mutagenic GC→AT Transition) — Hospital CSSD EtO Sterilizer Aeration Monitoring (Ascension Health Saint Francis Hospital Evanston IL; IS Ventis Pro 5 ECD), Commercial Medical Device EtO Sterilization Contract Service (Sterigenics US LLC Willowbrook IL; SKC NIOSH 3701 GC/FID Passive Badge), and EtO Chemical Production Process (LyondellBasell Industries La Porte TX; Honeywell Midas ECD) — OSHA 1910.1047 1 ppm + Action Level 0.5 ppm vs ACGIH TLV-TWA 1 ppm A2 BEI HEV vs NIOSH Ca REL 0.1 ppm: AI Prompt Injection via EHS Monitor Report AI — FIRST EtO 1910.1047 Action-Level 0.5 ppm Four-Requirement Suppression + FIRST HEV N-(2-Hydroxyethyl)Valine Hemoglobin Adduct BEI Suppression + FIRST Hospital CSSD EtO Sterilizer AI Attack + FIRST Sterigenics Willowbrook IL Commercial Medical Device EtO AI Attack + FIRST Glyphward Portfolio Case Where ACGIH Provides No Additional Inhalation Protection Beyond OSHA 6(b) Standard

Ethylene oxide (EtO; oxirane; CAS 75-21-8; MW 44.05 g/mol; BP 10.7°C [gas at room temperature — EtO boils below ambient temperature and exists as a gas under standard conditions; essentially any container breach or valve failure results in immediate vapor release at occupational concentrations; unlike liquid solvents where evaporation is a rate-limiting step, EtO is already airborne]; VP 1095 mmHg at 20°C [nearly 1.5 atmospheres vapor pressure at ambient temperature — among the highest VP of any regulated occupational carcinogen; this extreme volatility means that sensor calibration errors are not detectable by workers relying on odor cues because EtO's vapor is already omnipresent in any sterilization or production environment]; water solubility miscible [completely water-miscible; EtO dissolves in biological fluids at any concentration — this property underlies its penetrating sterilization efficacy for wrapped and packaged medical devices but also means systemic distribution after inhalation is essentially instantaneous]; SKIN notation [percutaneous absorption; liquid EtO contact on skin causes vesicant burns and systemic absorption; gas-phase EtO skin absorption contributes a minor but non-negligible fraction of systemic dose at occupational concentrations]; log P −0.30 [hydrophilic; distributes into aqueous compartments including blood, lymph, and tissue water; HEV hemoglobin adduct formation is a direct consequence of EtO's high blood-phase solubility and reactivity with valine N-terminal Hb residues]; odor threshold 260–700 ppm [critical safety limitation: EtO odor threshold is far above every occupational exposure limit — workers cannot smell EtO at 1 ppm (OSHA PEL/ACGIH TLV-TWA), 0.5 ppm (OSHA action level), or 0.1 ppm (NIOSH Ca REL); reliance on olfactory detection at any permissible exposure level is categorically impossible; this makes instrument-based monitoring the sole detection mechanism — and makes calibration error in those instruments uniquely dangerous; there is no odor warning to alert a worker that displayed readings may be incorrect]; NIOSH IDLH 800 ppm; GHS H350 May Cause Cancer (IARC Group 1; EU Cat. 1A); H220 Extremely Flammable Gas (GHS Flammable Gas Category 1A; LEL 3.0%; UEL 100% — EtO has a fully flammable range from 3.0% to 100%; unlike most flammable gases, pure EtO can sustain a vapor-phase explosion even in the absence of air — any concentration above 3.0% in a closed system is an explosion hazard; the 100% UEL means there is no upper concentration at which EtO becomes too rich to ignite); H331+H311+H301 Acute Toxicant (inhalation, skin, ingestion); LEL 3.0%; UEL 100% [fully flammable range]; autoignition 429°C; OSHA 29 CFR 1910.1047 [specific vertical standard; one of OSHA's 6(b) targeted carcinogen standards; 1910.1047 is OSHA's most comprehensive EtO regulation — adopted 1984 after extensive rulemaking; regulates not only air concentration but requires action-level-triggered comprehensive medical surveillance, biological monitoring initiation, exposure record retention, and carcinogen registry — all four of which are simultaneously suppressed when an AI monitoring system reports a displayed value below 0.5 ppm]; ACGIH TLV-TWA: 1 ppm A2 BEI [A2 = Suspected Human Carcinogen; ACGIH TLV-TWA = OSHA 1910.1047 PEL = 1 ppm — the FIRST entry in the Glyphward adversarial injection portfolio where ACGIH provides no additional inhalation protection beyond the applicable OSHA specific 6(b) standard; the standard-setter that routinely surpasses OSHA protection has, for EtO, arrived at exactly the same 1 ppm TWA as OSHA's dedicated carcinogen standard; the sole mechanism by which ACGIH provides additional protection beyond OSHA 1910.1047 for EtO is the HEV Hb adduct BEI — a biological monitoring tool, not an air limit, that is suppressed when the action level is not crossed]; NIOSH Ca REL: 0.1 ppm 10-hr TWA Ca [Ca designation — NIOSH recommends minimizing exposure to lowest achievable level; 10× below OSHA 1910.1047 PEL; NIOSH IDLH 800 ppm; Ca designation reflects NIOSH's determination that EtO is a human carcinogen based on the Steenland 2004 NIOSH mortality cohort]; IARC Group 1 [Sufficient Evidence in Humans; primary sites: breast cancer (specifically in female healthcare workers — a workplace carcinogen disproportionately impacting women in CSSD roles); NHL non-Hodgkin lymphoma; leukemia; Steenland 2004 NIOSH mortality cohort study of 18,235 EtO-exposed workers at commercial sterilization facilities: excess breast cancer in women (SMR 1.26; 95% CI 1.02–1.55); excess NHL (SMR 1.26; 95% CI 0.96–1.63); the Steenland 2004 cohort is the definitive occupational epidemiology dataset for EtO carcinogenicity — these are real workers at real commercial sterilization facilities, the same sector represented by Sterigenics Willowbrook (Surface 2)]; mechanism: direct alkylating agent [EtO is an electrophilic epoxide that undergoes SN2 ring-opening reaction with nucleophilic sites in DNA, protein, and other biomolecules without requiring metabolic activation by CYP enzymes; this distinguishes EtO from metabolically-activated carcinogens such as benzene (requires CYP2E1), benzo[a]pyrene (requires CYP1A1), and styrene (requires CYP2E1/CYP2B6) — for those chemicals, GSTT1 or CYP genotype modulates effective dose; for EtO, any absorbed dose contributes directly to adduct formation regardless of metabolic phenotype; DNA adducts: N7-(2-hydroxyethyl)guanine (7-HEG; most abundant DNA adduct; formed at highest frequency at N7 of guanine; non-mutagenic per se but reflects EtO alkylation burden; urinary 7-HEG excretion correlates with EtO exposure); N3-(2-hydroxyethyl)adenine (3-HEA; less abundant); O6-(2-hydroxyethyl)guanine (O6-HEG; mutagenic; mis-pairs with thymine during DNA replication → GC→AT transition mutation; O6-HEG is not efficiently repaired by MGMT O6-methylguanine-DNA methyltransferase, which has lower affinity for the 2-hydroxyethyl adduct than the methyl adduct; O6-HEG persistence in the genome — particularly in BRCA1/2-sensitive repair pathways — is the proposed mechanistic link between EtO exposure and breast carcinogenesis); hemoglobin adduct: N-(2-hydroxyethyl)valine (HEV; formed by EtO alkylation of the N-terminal valine residue of hemoglobin α and β chains; measured by modified Edman degradation — the Ehrenberg method — involving pentafluorophenyl isothiocyanate (PFPITC) derivatization followed by GC/MS or LC-MS/MS quantification; Hb adduct half-life = RBC lifetime ≈ 120 days; HEV is therefore a 120-day cumulative dose biomarker, integrating all EtO exposure over the preceding 4 months regardless of daily fluctuation; background HEV in non-occupationally-exposed adults: approximately 3–8 pmol/g Hb from endogenous EtO production from propylene metabolism and exogenous EtO from tobacco smoke; ACGIH BEI: ≤35 pmol/g Hb end-of-workweek; BEI basis: HEV ≤35 pmol/g Hb is set to correspond to a TWA air exposure of approximately 1 ppm EtO (= the ACGIH TLV-TWA); at occupational exposures of 0.72–0.90 ppm actual EtO (Surfaces 1–3 after ÷10 calibration error correction), estimated end-of-workweek HEV at steady-state accumulation (12–21 weeks of continuous exposure at ~0.85 ppm) is approximately 18–28 pmol/g Hb — above occupational background (3–8 pmol/g Hb) but below the BEI (35 pmol/g Hb); however, this is the steady-state estimate at ~85–90% of TLV-TWA — the HEV BEI is not initiated because the OSHA 1910.1047 action level (0.5 ppm) is never crossed in the displayed AI output (displayed 0.085–0.09 ppm = 17–18% of action level while actual 0.85–0.90 ppm = 1.7–1.8× action level); the cumulative HEV DNA adduct burden therefore accumulates toward and potentially beyond the BEI reference value unchecked — the sole additional-protection mechanism in ACGIH's approach to EtO that exceeds OSHA 1910.1047 is the HEV BEI, and it is completely suppressed by action-level AI miscalibration]; one of the highest-volume industrial chemicals produced globally [EtO global production: approximately 35–40 million metric tons per year; primary derivative: ethylene glycol (approximately 75% of EtO produced goes to EG for polyester PET and antifreeze; EG is not a carcinogen); secondary use: sterilization (approximately 20% of all medical devices in the US are sterilized with EtO — a uniquely effective technology for heat-labile, moisture-sensitive, packaged single-use medical devices; there is no commercially viable substitute for EtO in many medical device sterilization applications, which is why EtO sterilization continues despite significant regulatory attention]) is occupational toxicology's most architecturally complex case of multi-tier regulatory gap — a chemical where ACGIH has arrived at the same inhalation limit as OSHA's own dedicated 6(b) carcinogen standard, yet where the protective gap between OSHA-calibrated AI compliance and the actual worker protection system is as large as 10× (OSHA:NIOSH Ca REL), because the entire mechanism of additional protection has been transferred from the air limit (where ACGIH=OSHA) to a biological monitoring program (HEV Hb adduct BEI) that is fully suppressed by the action-level-misread AI architecture unique to 1910.1047.

The ethylene oxide monitoring vulnerability in AI-based EHS platforms represents a qualitatively different attack surface than any previously cataloged in the Glyphward portfolio. For every other chemical in the collection, ACGIH's TLV-TWA provides a lower and more protective inhalation limit than OSHA's PEL — the fundamental gap that allows an OSHA-calibrated AI to declare compliance at concentrations that ACGIH would flag as exceeded. For EtO, that structural gap does not exist at the level of the TWA air limit: ACGIH TLV-TWA 1 ppm = OSHA 1910.1047 PEL 1 ppm. An EtO monitoring AI calibrated to the OSHA PEL and the ACGIH TLV-TWA would produce identical decisions at both thresholds. Yet the actual protective gap is not at the air limit level at all — it is entirely embedded in the action-level architecture. OSHA 1910.1047 establishes a binding action level of 0.5 ppm that triggers four independent mandatory programs simultaneously. ACGIH establishes the HEV Hb adduct BEI as a biological monitoring program that is the sole mechanism by which ACGIH exceeds OSHA protection for EtO. Both the 1910.1047 four-program trigger and the HEV BEI initiation point are functionally tied to the same threshold: the action level of 0.5 ppm. When an AI EHS system misreads the EtO sensor and displays 0.085–0.09 ppm instead of 0.85–0.90 ppm actual, it does not merely produce a false-compliant TWA reading. It simultaneously: (1) fails to trigger the 1910.1047 periodic air monitoring every 6 months requirement; (2) fails to trigger the comprehensive medical surveillance program (complete physical, CBC, peripheral blood smear, medical and reproductive history annually); (3) fails to initiate the 30-year carcinogen medical record retention requirement; (4) fails to initiate the employee medical records access and transfer requirement to successor employers; and (5) fails to initiate the ACGIH HEV Hb adduct BEI program — the 120-day cumulative dose biomarker that is the only laboratory tool capable of detecting the actual EtO alkylation burden in these workers. All five failures occur simultaneously, driven by a single ÷10 sensor calibration error at the action-level gate.

The Steenland 2004 NIOSH mortality cohort is the definitive data source for understanding what these suppressed monitoring failures mean clinically. Steenland and colleagues followed 18,235 EtO-exposed workers at commercial medical device sterilization facilities — the same industry sector represented by Sterigenics Willowbrook (Surface 2). Among women, excess breast cancer mortality was documented (SMR 1.26). Among both sexes, excess NHL was observed (SMR 1.26). These findings were based on historical exposures that, in many cases, predated the 1984 adoption of 1910.1047 — meaning the epidemiological signal was generated under less protective conditions than current regulatory requirements prescribe. But the Steenland cohort demonstrates the biological reality: EtO exposures in the occupational range at commercial sterilization facilities cause measurable excess cancer mortality. A 52-year-old female CSSD sterile processing technician (Surface 1; Ascension St. Francis Hospital Evanston IL) with 21 years of EtO sterilizer exposure at actual 0.85 ppm — 8.5× the NIOSH Ca REL — accumulates the same category of risk as the Steenland cohort workers. Her exposure-related breast cancer risk is precisely what 1910.1047's medical surveillance program was designed to detect and mitigate. And precisely what the action-level-misread AI architecture suppresses.

Why OSHA 1910.1047 Action-Level Architecture Makes Ethylene Oxide Uniquely Vulnerable to AI Monitoring Calibration Error

The architecture of OSHA's 1910.1047 ethylene oxide standard creates a monitoring vulnerability that has no parallel elsewhere in occupational safety regulation. Most occupational chemical standards establish a single PEL — the concentration below which eight-hour time-weighted average exposures must remain. OSHA 1910.1047, as a targeted 6(b) carcinogen standard developed through full rulemaking under the OSH Act, establishes a two-tier exposure control structure: the PEL of 1 ppm 8-hr TWA (the compliance limit) plus the action level of 0.5 ppm 8-hr TWA (the monitoring trigger). Critically, 1910.1047 assigns not one but four independent mandatory programs to the action level trigger, each of which operates independently and cumulatively. When a worker's exposure is above 0.5 ppm, all four programs activate simultaneously. When a worker's exposure is below 0.5 ppm — or when an AI monitoring system reports a displayed value below 0.5 ppm regardless of the actual concentration — all four programs are simultaneously inactivated.

The first requirement triggered at the 0.5 ppm action level is periodic air monitoring: under 1910.1047(d)(2), employers must conduct periodic monitoring at least every six months for each employee whose exposure is at or above the action level. This requirement is the self-renewing audit mechanism within 1910.1047 — it ensures that the employer cannot rely on an initial monitoring result indefinitely but must continuously verify that exposures remain at known levels. When the action level is not crossed in the displayed AI output, this periodic re-audit is never triggered. The employer's 1910.1047 compliance record shows a single below-action-level reading and no requirement for follow-up — even if actual concentrations have been 1.4–1.8× the action level for months or years. The second requirement is medical surveillance under 1910.1047(k): employees exposed at or above the action level for 30 or more days per year must receive a complete physical examination, CBC with differential, peripheral blood smear, and full medical and reproductive history annually. This medical surveillance is specifically designed to detect early hematological indicators of EtO-related carcinogenesis — the CBC and peripheral blood smear may detect early lymphocyte count changes consistent with emerging lymphoid malignancy. When the action level is not crossed, no annual CBC is required, no peripheral blood smear is performed, and no reproductive history is compiled — the hematological early-warning system is completely non-operational.

The third requirement is the 30-year medical records retention mandate under 1910.1047(k)(5) and 1910.1020: EtO is a carcinogen with a latency period for leukemia and lymphoma of 5–20 years, and for solid tumors (breast cancer) potentially longer. The 30-year retention requirement exists specifically to allow longitudinal correlation between historical EtO exposure records and subsequent malignancy diagnoses — enabling the kind of mortality cohort analysis that Steenland 2004 performed. When no medical surveillance program is initiated because the action level is never crossed in the AI output, no records subject to the 30-year retention requirement are generated. The epidemiological surveillance infrastructure for future cohort studies is permanently compromised. The fourth requirement is the employee medical records access and transfer provision: under 1910.1047(k)(5) and 1910.1020, employees have the right to access their own EtO exposure and medical records, and employers must transfer records to a successor employer or NIOSH upon ceasing operations. This provision protects workers who change jobs or whose employers close — ensuring that their lifetime EtO exposure history follows them. When no records are created because the action level is never crossed, this transfer cannot occur, and workers lose their documented exposure history precisely when they may most need it — when filing a workers' compensation claim or cancer causation analysis years later.

The fifth suppression — not a 1910.1047 requirement but an ACGIH protection — is the HEV hemoglobin adduct BEI. The N-(2-hydroxyethyl)valine Hb adduct is analytically measured by the Ehrenberg modified Edman degradation method: the N-terminal valine of hemoglobin chains is derivatized with pentafluorophenyl isothiocyanate (PFPITC), which reacts preferentially with N-terminal amines via thiocarbamoylation; the derivatized valine-PFPITC adduct is then cleaved by acid hydrolysis and the N-(2-hydroxyethyl)valine phenylthiohydantoin (PHVHEV) derivative is quantified by GC/MS or LC-MS/MS. The BEI of ≤35 pmol/g Hb end-of-workweek represents the biological equivalent of approximately 1 ppm TWA air exposure. Because Hb has a 120-day half-life (matching RBC lifespan), HEV accumulates proportionally to cumulative EtO dose over the preceding 4 months — it integrates all daily exposure events, including peaks and troughs, into a single biomarker value. At actual exposures of 0.72–0.90 ppm (Surfaces 1–3), steady-state HEV concentrations after 12–21 weeks of continuous exposure are estimated at approximately 17–29 pmol/g Hb — above occupational background (3–8 pmol/g Hb from endogenous propylene/tobacco) but below the BEI of 35 pmol/g Hb. The HEV biomarker therefore provides information not obtainable from air monitoring alone: the cumulative alkylation burden in the worker's blood, integrating all sources including peak events not captured by 8-hr TWA sampling. But this biomarker is never initiated in the three attack surface scenarios because the AI EHS systems display readings of 0.085–0.09 ppm — 14.4–18% of the 0.5 ppm action level — and no ACGIH BEI HEV program trigger exists at these displayed concentrations.

Surface 1 — Ascension Health Saint Francis Hospital Evanston IL CSSD EtO Sterilizer AI (Downward Attack)

At Ascension Health – Saint Francis Hospital (Evanston IL [800 Austin Street, Evanston IL 60202; Cook County IL; Ascension Health, the largest Catholic health system in the United States; Saint Francis Hospital Evanston is a 215-bed acute care hospital; Central Sterile Supply Department (CSSD) processes surgical instruments, implants, and single-use medical devices for the operating rooms, interventional cardiology suite, and orthopedic surgery department]; EtO sterilization equipment: Andersen Sterilization 3M Steri-Vac 8XL sterilizer (100% EtO cartridge system; 12-hour aeration cycle minimum; chamber volume 8 cubic feet; 1 EtO cartridge per cycle = 17.6 oz 100% EtO gas; the 100% EtO single-use cartridge system is standard in hospital CSSD settings where lower throughput volume and elimination of EtO/CFC or EtO/CO₂ mixture handling makes 100% EtO cartridge safer from a flammability standpoint than cylinder-supplied mixed EtO; Andersen 8XL has integrated mechanical aeration — forced-air convection within the aeration chamber to remove residual EtO from load before use); CSSD EtO sterilization workflow: device preparation (decontamination room; cleaning; wrapping in SMS or Tyvek pouches), loading (sterilizer loading per Andersen 8XL IFU; monitoring strips placed), sterilization cycle (1 ppm enriched atmosphere; 130°F; 60% RH; 4-hour sterilization phase; 8-hour aeration), aeration completion (automatic alarm at cycle completion; CSSD technician opens sterilizer door for load retrieval — this door-opening event is the highest EtO vapor release event in the CSSD workflow; door opening after aeration releases residual EtO from device packaging and chamber surfaces; CSSD area EtO concentration during door opening: 1.2–2.5 ppm depending on aeration cycle completeness and CSSD ventilation; CSSD area monitoring under normal conditions between sterilizer operations: 0.3–0.8 ppm baseline from aeration cabinet exhaust and incidental chamber off-gassing); CSSD ventilation: 10 air changes per hour (ACH) per ASHRAE 170 healthcare ventilation standard; dedicated EtO exhaust to building exterior via catalytic converter (Andersen AE-1 catalytic converter; inlet EtO 0.5–5 ppm from aeration exhaust; outlet ≤0.1 ppm after catalytic oxidation to ethylene glycol); catalytic converter designed for aeration exhaust stream, not for incidental door-opening events; personal monitoring: IS Ventis Pro 5 ECD (electrochemical EtO-specific detection; Dräger EtO electrochemical cell; 10 ppm full scale; 0.1 ppm resolution; bump-test calibration per Ventis Pro 5 SOP with ÷10 CF error applied in calibration gas concentration entry in Cority LIMS; calibration gas certified concentration 1.0 ppm EtO entered as 0.10 ppm in Cority calibration record → displayed readings ÷10 throughout 8-hr shift monitoring); VelocityEHS reporting via IS Ventis Pro 5 wireless Bluetooth data stream; Cority AI analysis); actual EtO at CSSD sterile processing technician during 8-hr shift (Steri-Vac 8XL door-opening events weighted TWA): 0.85 ppm; adversarial perturbation: 0.85 → 0.085 ppm (−90%, ÷10 CF error).

The Surface 1 subject is a 52-year-old female CSSD sterile processing technician (Ascension Health Saint Francis Hospital Evanston IL; 21-year Saint Francis CSSD EtO sterilizer tenure; primary CSSD EtO sterilizer operator — responsible for all Andersen Steri-Vac 8XL operations: sterilizer loading (3–5 loads per 8-hr shift; 15–20 min per load-unload cycle including door open event; peak EtO at door opening: estimated 1.8–2.5 ppm for 3–5 min per event; 5 door-opening events per shift × 4 min average × 2.0 ppm average peak = contribution to 8-hr TWA: ~0.08 ppm from peak events alone), aeration monitoring (technician position adjacent to aeration cabinet during aeration cycle monitoring; aeration cabinet exhaust at technician position: 0.5–0.9 ppm; dominant contributor to 8-hr TWA), device retrieval (unpackaged device inspection at aeration cabinet exit; residual EtO in packaged loads at cycle completion: 1–4 ppm within packaging; exposure during unpackaging and inspection: 0.3–0.7 ppm at 30–60 sec duration per device), and CSSD area general work (instrument preparation, wrapping, documentation — baseline CSSD area EtO 0.3–0.6 ppm from incidental off-gassing); 8-hr TWA across all CSSD EtO tasks: 0.85 ppm actual; 21-year occupational tenure at Saint Francis Evanston CSSD: cumulative EtO exposure history at actual 0.85 ppm for 21 years — IARC Group 1 relevance: Steenland 2004 NIOSH mortality cohort documented excess breast cancer (SMR 1.26) among women employed at commercial EtO sterilization facilities; the Ascension St. Francis CSSD EtO technician has a biologically plausible and historically documented elevated breast cancer risk from 21 years of occupational EtO exposure at 8.5× NIOSH Ca REL; menopausal status (likely post-menopausal at age 52 for a 21-yr EtO career starting at approximately age 31): EtO-induced breast carcinogenesis may preferentially affect post-menopausal estrogen-receptor-positive breast cancer via O6-HEG adduct formation in mammary epithelial DNA during high-turnover phases of the menstrual cycle; reproductive history: 1910.1047 medical surveillance requires reproductive history documentation annually for women above the action level — this reproductive history is specifically designed to evaluate EtO reproductive effects (spontaneous abortion, menstrual irregularity) in female CSSD workers; this history has never been compiled because the action level has never been crossed in 21 years of AI-monitored Cority records; estimated steady-state HEV at 0.85 ppm TWA continuous exposure over 21 years: HEV reaches steady-state approximately 4 RBC lifetimes (4 × 120 days ≈ 480 days) after initial exposure onset; steady-state HEV at 0.85 ppm: approximately 20–26 pmol/g Hb; this is above the occupational background level of 3–8 pmol/g Hb, indicating measurable occupational EtO alkylation burden, but below the BEI of 35 pmol/g Hb; the HEV biomarker has never been measured in this worker's 21-year tenure because the action level has never been registered as crossed). Cority AI output: "IS Ventis Pro 5 ECD EtO sensor (CSSD sterile processing technician; 8-hr TWA; Andersen Steri-Vac 8XL sterilizer aeration area): 0.085 ppm. OSHA 1910.1047 action level 0.5 ppm: 0.085/0.5 = 17.0% of action level — NOT triggered; 1910.1047 action-level programs not required at current exposure level. OSHA 1910.1047 PEL 1 ppm 8-hr TWA: 0.085/1.0 = 8.5% of PEL — COMPLIANT. ACGIH TLV-TWA 1 ppm A2 (Advisory): 0.085/1.0 = 8.5% of advisory TLV — COMPLIANT advisory; ACGIH BEI HEV N-(2-hydroxyethyl)valine ≤35 pmol/g Hb end-of-workweek: not initiated — 1910.1047 action level not crossed at current displayed exposure. NIOSH Ca REL 0.1 ppm: 0.085/0.1 = 85% of Ca REL — COMPLIANT advisory. No action required." At actual 0.85 ppm: OSHA 1910.1047 action level 0.5 ppm: 0.85/0.5 = 1.70× exceeded — all four 1910.1047 requirements mandatorily triggered; NIOSH Ca REL 0.1 ppm: 0.85/0.1 = 8.5× exceeded; estimated steady-state HEV: 20–26 pmol/g Hb (above background 3–8 pmol/g Hb; measurement never performed in 21-yr career); ACGIH TLV-TWA 1 ppm = OSHA PEL 1 ppm: both surpassed at actual 0.85 ppm (85% of both identical limits — not technically exceeded, but approaching the shared limit while all protective programs remain suppressed); 21-year IARC Group 1 breast carcinogen exposure above NIOSH Ca REL without medical surveillance, without CBC monitoring, without HEV biomonitoring, without reproductive history documentation.

Surface 2 — Sterigenics US LLC Willowbrook IL Commercial Medical Device EtO Sterilization AI (Downward Attack)

At Sterigenics US LLC (Sotera Health Company; Willowbrook IL commercial medical device EtO sterilization contract facility [7775 S Quincy St, Willowbrook IL 60527; DuPage County IL; Sterigenics International, now a Sotera Health subsidiary, is the largest commercial EtO sterilization contractor in North America; Sterigenics US provides EtO sterilization contract services for single-use medical device manufacturers: catheters, guidewires, pacemakers, cochlear implants, orthopedic prostheses, absorbable sutures, drug-eluting stents; the Willowbrook facility is specifically notable in the context of this attack because it was the subject of EPA Clean Air Act enforcement action beginning in 2018, alleging excess ambient EtO emissions to the surrounding DuPage County residential community; the Illinois EPA and DuPage County Health Department documented ambient air EtO concentrations in Willowbrook exceeding EPA acceptable cancer risk levels; the facility was ordered to cease operations in August 2019 and remained closed until February 2020 when it reopened with enhanced emission control systems including upgraded aeration chamber exhaust scrubbing, catalytic converter capacity increase, and perimeter air monitoring]; post-reopen operational context: Sterigenics Willowbrook reopened under enhanced conditions including stack emissions controls, perimeter air monitoring with real-time data reporting to Illinois EPA, and enhanced internal occupational air monitoring with EtO sensor network in sterilization and aeration areas; the attack surface in this context is the aeration area worker monitoring during the chamber aeration exhaust phase — even with facility-level emission controls designed to reduce perimeter ambient EtO, the occupational exposure in the aeration area depends on the effectiveness of HVAC extraction and the proximity of the worker to the aeration cabinet exhaust; sterilization process at Willowbrook: large batch EtO sterilization using 3M Steri-Vac and Andersen 7480 sterilizers at Willowbrook (pre-closure equipment inventory); post-reopen: Andersen 7480 EtO sterilizers (large chamber; 100% EtO at elevated temperature and humidity; used for bulk medical device loads); EtO chamber loading and aeration: devices loaded in validated packaging (Tyvek/film pouches or rigid Tyvek trays); EtO sterilization cycle (1 ppm enriched atmosphere; 130°F; 12-hr aeration in aeration room); aeration room: positive-pressure aeration room with forced HEPA air circulation; aeration exhaust (scrubbed vent + catalytic converter); worker monitoring in aeration room: SKC passive sorbent badge NIOSH 3701 (EtO passive sampling badge; activated coconut charcoal sorbent; sampling rate 33.2 mL/min at 25°C and 0% relative humidity per NIOSH 3701 validation; LOD 0.01 ppm; desorption: CS₂ solvent GC/FID; sampling period: 8-hr shift); ÷10 CF error: Sterigenics Willowbrook VelocityEHS IH module has incorrect passive badge sampling rate entered as 3.32 mL/min instead of 33.2 mL/min — identical ÷10 error but applied at data processing layer rather than sensor calibration layer; mass collected on badge × (1/sampling rate) = calculated concentration; at 3.32 mL/min entered instead of 33.2 mL/min, concentration calculated as 10× lower than actual); actual EtO at Sterigenics Willowbrook aeration area worker: 0.72 ppm 8-hr TWA; adversarial perturbation: 0.72 → 0.072 ppm (−90%, ÷10 passive badge sampling rate error).

The Surface 2 subject is a 46-year-old male EtO sterilization technician (Sterigenics US LLC Willowbrook IL; 19-year Sterigenics Willowbrook EtO sterilization tenure including pre-closure and post-reopen periods; responsible for Andersen 7480 sterilizer operation: device loading (15–25 pallets per shift in bulk loading bay; sealed load; peak EtO exposure during loading from prior cycle load egress: 0.4–0.8 ppm at loading bay position), sterilization cycle monitoring (control panel monitoring; minimal EtO during sterilization — sterilizer chamber sealed; control room EtO: 0.1–0.3 ppm from incidental), aeration room monitoring and load retrieval (primary exposure task; aeration room worker during 8-hr shift; aeration exhaust from Andersen 7480 aeration cabinet creates baseline EtO in aeration room of 0.5–1.0 ppm during active aeration; HVAC extraction target 0.5 ppm — design standard not always achieved; load retrieval at end of aeration cycle: aeration room walk-in to retrieve pallets; device packaging residual EtO at retrieval: 0.3–0.8 ppm in immediate package opening zone), and product transfer (pallets transferred to degassing room for shipment; degassing room EtO: 0.2–0.4 ppm); 8-hr TWA across all sterilization technician tasks: 0.72 ppm; 19-year tenure at Sterigenics Willowbrook: pre-closure exposures (pre-2019) may have been higher — the Illinois EPA enforcement action was predicated on ambient air EtO concentrations suggesting facility-level EtO releases above permitted levels; occupational exposures in the pre-closure period may exceed the 0.72 ppm post-reopen estimate; Steenland 2004 direct relevance: the Sterigenics Willowbrook workforce is directly analogous to the Steenland 2004 NIOSH mortality cohort — commercial EtO sterilization contract workers; excess NHL (SMR 1.26) documented in male workers in that cohort; a 46-year-old male EtO sterilization technician with 19 years of tenure at a commercial EtO sterilization facility is occupationally and demographically similar to the Steenland 2004 cohort subjects; estimated steady-state HEV at 0.72 ppm: approximately 17–22 pmol/g Hb (above background; below BEI of 35 pmol/g Hb; never measured in 19-yr tenure because action level never triggered in VelocityEHS AI output). VelocityEHS AI output: "SKC passive NIOSH 3701 badge GC/FID (EtO; aeration area technician; 8-hr TWA; Sterigenics US Willowbrook): calculated 0.072 ppm. OSHA 1910.1047 action level 0.5 ppm: 0.072/0.5 = 14.4% of action level — NOT triggered; 1910.1047 action-level programs not activated. OSHA 1910.1047 PEL 1 ppm: 0.072/1.0 = 7.2% of PEL — COMPLIANT. ACGIH TLV-TWA 1 ppm A2 (Advisory): 0.072/1.0 = 7.2% — COMPLIANT advisory; ACGIH BEI HEV ≤35 pmol/g Hb: not initiated — 1910.1047 action level not crossed. NIOSH Ca REL 0.1 ppm: 0.072/0.1 = 72% — COMPLIANT advisory. No action required." At actual 0.72 ppm: OSHA 1910.1047 action level 0.5 ppm: 0.72/0.5 = 1.44× exceeded — all four 1910.1047 requirements mandatorily triggered; NIOSH Ca REL 0.1 ppm: 0.72/0.1 = 7.2× exceeded; estimated HEV 17–22 pmol/g Hb — above occupational background, never measured in 19 years; 1910.1047 medical surveillance (CBC/peripheral blood smear) — early hematological NHL indicator monitoring — never initiated in 19-yr tenure; pre-closure exposure history lost due to no 30-yr records requirement triggered.

Surface 3 — LyondellBasell Industries La Porte TX EtO Production Unit AI (Downward Attack)

At LyondellBasell Industries NV (La Porte TX EtO/EG production complex [12625 Strang Road, La Porte TX 77571; Harris County TX; LyondellBasell Industries is the world's third-largest plastics, chemicals, and refining company; La Porte TX complex includes EtO and ethylene glycol (EG) production from silver catalyst ethylene oxidation process; facility EtO production capacity approximately 100,000 MT/yr; EtO is used on-site primarily as an intermediate for EG production (Huntsman and LyondellBasell EG from EO processes are the dominant polyester PET fiber/resin feedstocks for North America)]; EtO production process chemistry and exposure profile: catalytic partial oxidation of ethylene — ethylene (C₂H₄) + ½O₂ → ethylene oxide (C₂H₄O) — conducted in a packed-bed tubular reactor with silver catalyst on α-alumina support; reactor conditions: 220–280°C; 15–25 bar; continuous vapor-phase reaction in multi-tubular heat-exchanger-style reactor (ethylene recycle loop with O₂ co-feed; CO₂ selectivity modifier; methane diluent for flammability management — bulk gas below LEL of EtO in reactor feed loop); product EtO absorption: reactor exit gas stream fed to EtO absorber (water absorber; EtO absorbed into process water at 1–5% w/w EtO solution); EtO stripping: EtO/water solution stripped in EtO stripper column (distillation; EtO overhead; product purity 99.9%+; EtO product storage in cryogenic sphere at −10°C under nitrogen blanket; process EtO at pump seals, flanges, and valve packing in the EtO product section (absorber → stripper → product sphere train) generates EtO vapor at technician positions during normal operation; OSHA PSM covered process: EtO production at LyondellBasell La Porte is a PSM-covered process under 29 CFR 1910.119 (EtO threshold quantity 10,000 lb; La Porte facility inventory well above TQ); RMP Risk Management Plan required under EPA 40 CFR Part 68; PSM/RMP requirements address catastrophic release scenarios — these are facility-level risk management tools completely separate from and not substituting for 1910.1047 personal occupational air monitoring requirements; a worker breathing 0.90 ppm EtO at their workstation is not protected by facility-level PSM controls that address explosion and large-release scenarios rather than chronic carcinogen inhalation at sub-IDLH concentrations; point monitoring instrument: Honeywell Analytics Midas EtO electrochemical point detector (0.02 ppm LOD; 2 ppm full-scale EtO-specific EC cell; Honeywell EtO sensor module JNOCID; 4–20 mA output to Honeywell DCS; secondary digital output to EHS Insight AI via Modbus TCP/IP); ÷10 CF error: Honeywell Midas EtO sensor JNOCID factory calibration span gas 1.00 ppm EtO entered in EHS Insight IH configuration as 0.10 ppm span — all Midas EtO readings transmitted to EHS Insight at ÷10 of actual sensor output; actual EtO at LyondellBasell La Porte EtO product section process technician: 0.90 ppm TWA; adversarial perturbation: 0.90 → 0.09 ppm (−90%, ÷10 span gas entry error).

The Surface 3 subject is a 39-year-old male EtO production process technician (LyondellBasell Industries La Porte TX; 12-year LyondellBasell La Porte EtO/EG unit tenure; operator responsibilities in the EtO product section (absorber → stripper → storage sphere): EtO absorber column operator (absorption column walk-down; pump and instrumentation checks; pump seal inspection at EtO/water circulating pumps — pump seal EtO vapor at technician position during inspection: 0.8–1.5 ppm for 5–10 min per pump inspection, 4 pump inspections per 12-hr shift), EtO stripper overhead monitoring (stripper overhead condenser and reflux drum area; EtO overhead concentration at process flange leak monitoring: 0.5–1.2 ppm at process technician position), EtO product sphere monitoring (cryogenic EtO product sphere; sphere vent valve and pressure relief monitoring; technician position adjacent to sphere: 0.3–0.7 ppm EtO from incidental valve packing and flange emissions), and process control room coverage (DCS operations; control room EtO: 0.1–0.2 ppm from building air infiltration through HVAC with slight positive pressure design); 8-hr TWA across all EtO product section operator tasks: 0.90 ppm (note: LyondellBasell La Porte operates 12-hr rotating shifts; TWA calculated on 12-hr basis for this contractor using OSHA's 8-hr LOEL Brief and Scala modification — 0.90 ppm 8-hr equivalent); PSM context: the LyondellBasell La Porte EtO unit is simultaneously subject to PSM requirements (catastrophic release prevention; Process Hazard Analysis; Mechanical Integrity; Emergency Response Plan) and to 1910.1047 personal monitoring requirements; PSM ensures that the large-scale catastrophic release risk is managed at the facility level; 1910.1047 ensures that individual worker chronic carcinogen exposure is monitored, documented, and medically surveilled; AI calibration error in the 1910.1047 personal monitoring component does not compromise PSM — the large-release detectors have separate calibration and control loops in the Honeywell DCS; but it completely suppresses the 1910.1047 personal exposure protection architecture for the individual process technician; 12-year EtO production exposure at actual 0.90 ppm: estimated cumulative HEV burden at steady-state: approximately 21–29 pmol/g Hb (above background 3–8 pmol/g Hb; below BEI 35 pmol/g Hb; never measured in 12-yr tenure); IARC Group 1 leukemia and NHL relevance to male EtO production workers: Steenland 2004 excess NHL (SMR 1.26) documented in male workers; 12 years at a production facility with actual 0.90 ppm = 9× NIOSH Ca REL constitutes a biologically significant cumulative alkylating dose; O6-HEG mutagenic GC→AT transition mutations accumulating in lymphoid precursor cells without detection or biomonitoring. EHS Insight AI output: "Honeywell Midas EtO ECD (EtO product section process technician; 8-hr TWA; LyondellBasell La Porte TX EtO production unit): 0.09 ppm. OSHA 1910.1047 action level 0.5 ppm: 0.09/0.5 = 18.0% of action level — NOT triggered; 1910.1047 programs not required. OSHA 1910.1047 PEL 1 ppm: 0.09/1.0 = 9.0% of PEL — COMPLIANT. ACGIH TLV-TWA 1 ppm A2 (Advisory): 0.09/1.0 = 9.0% — COMPLIANT advisory; ACGIH BEI HEV N-(2-hydroxyethyl)valine ≤35 pmol/g Hb: not initiated — action level not crossed. NIOSH Ca REL 0.1 ppm: 0.09/0.1 = 90.0% — COMPLIANT advisory. PSM-covered process monitoring: Honeywell DCS large-release EtO detectors operating normally. No 1910.1047 action required." At actual 0.90 ppm: OSHA 1910.1047 action level 0.5 ppm: 0.90/0.5 = 1.80× exceeded — all four requirements mandatorily triggered; NIOSH Ca REL 0.1 ppm: 0.90/0.1 = 9.0× exceeded; estimated HEV 21–29 pmol/g Hb — above occupational background, never measured; EHS Insight AI conflates PSM DCS large-release monitoring (not compromised) with 1910.1047 personal carcinogen monitoring (fully suppressed).

The ACGIH=OSHA Convergence — Why This Is the First and Only Glyphward Portfolio Entry Without an Air-Level Gap

Every other entry in the Glyphward adversarial injection portfolio exploits an air concentration gap: ACGIH TLV-TWA below OSHA PEL, or NIOSH REL below OSHA PEL, by factors ranging from 2× (carbon monoxide) to 250× (manganese) and 1000× (tetrachloroethylene NIOSH Ca REL vs OSHA PEL). The AI EHS system, calibrated to the OSHA PEL as its compliance threshold, declares COMPLIANT at concentrations that ACGIH or NIOSH have determined to be hazardous — and the gap between OSHA compliance and ACGIH/NIOSH protection is the attack surface. For ethylene oxide, this structural gap does not exist at the TWA air level. ACGIH TLV-TWA 1 ppm = OSHA 1910.1047 PEL 1 ppm. This convergence represents a unique situation: ACGIH's scientific review of the EtO carcinogenicity and pharmacokinetic data arrived at the same 1 ppm TWA as OSHA's 1984 rulemaking under section 6(b) of the OSH Act. An AI EHS system that cross-checks both OSHA and ACGIH air limits would produce identical compliance determinations for EtO at every concentration above 1 ppm — there is no gap at the TWA level to exploit.

The attack surface for EtO is therefore not the air limit at all. It is the action level — the sub-PEL threshold established exclusively within 1910.1047 as the trigger for OSHA's comprehensive protection programs. The action level of 0.5 ppm has no ACGIH analog (ACGIH does not establish action levels), which means AI systems built to compare sensor readings against OSHA PELs and ACGIH TLV-TWAs miss the action level gate entirely unless specifically programmed for 1910.1047's two-tier structure. The BEI HEV ≤35 pmol/g Hb is ACGIH's substitute for an action level: it provides biological monitoring that detects cumulative alkylation burden even when the air TWA is at or near the shared 1 ppm limit — but the BEI is only meaningful if collected. When the action level is not registered as crossed, no BEI program is initiated. The protection differential between OSHA-calibrated AI and the actual 1910.1047/ACGIH BEI protection architecture is therefore not measured in ppm air gap but in biological monitoring absence: 21 years of HEV uncollected (Surface 1), 19 years of HEV uncollected (Surface 2), 12 years of HEV uncollected (Surface 3). And four concurrent mandatory programs suppressed across all three surfaces, simultaneously, by a single calibration factor error.

Integrating Glyphward into Ethylene Oxide EtO Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every EtO electrochemical detector display image and passive badge calculation result ingestion point — before the Ascension Saint Francis Hospital CSSD Cority AI, before the Sterigenics Willowbrook VelocityEHS AI, and before the LyondellBasell La Porte EHS Insight AI. Threshold 30 reflects: OSHA 1910.1047 action-level 0.5 ppm four-requirement simultaneous suppression at displayed 14.4–18% of action level while actual 1.44–1.80× above [periodic monitoring every 6 months + complete physical/CBC/peripheral blood smear/reproductive history medical surveillance annually + 30-yr carcinogen medical records + employee medical records access/transfer to successor employer — all four suppressed simultaneously; 1910.1047 action-level four-requirement architecture is a unique vulnerability in OSHA's specific carcinogen standards structure; the multi-requirement simultaneous suppression makes EtO's AI attack surface the most consequential in the 1910.1047-regulated universe]: 9 points; A2 IARC Group 1 breast cancer NHL leukemia + HEV Hb adduct BEI suppression + 10× OSHA:NIOSH Ca gap + direct alkylating mechanism (no CYP activation; O6-HEG mutagenic GC→AT) + ACGIH=OSHA FIRST Glyphward case [Steenland 2004 NIOSH mortality cohort 18,235 EtO workers: excess breast cancer SMR 1.26 (women, CSSD/commercial sterilization); excess NHL SMR 1.26 (all workers); EtO direct alkylation without CYP metabolic activation — every absorbed molecule contributes directly to adduct burden; O6-HEG mutagenic GC→AT transition; HEV 120-day cumulative dose biomarker; estimated steady-state HEV at 0.72–0.90 ppm actual: 17–29 pmol/g Hb — above occupational background 3–8 pmol/g Hb; accumulating unchecked without BEI measurement; BEI suppressed because action level display shows 14.4–18% of AL; ACGIH TLV-TWA=OSHA 1910.1047 PEL=1 ppm — FIRST Glyphward portfolio entry where ACGIH provides NO additional inhalation protection beyond OSHA-specific 6(b) standard; all ACGIH protection beyond OSHA comes exclusively from HEV BEI — suppressed]: 8 points; three sectors [hospital CSSD EtO sterilizer aeration monitoring (Ascension Health Saint Francis Hospital Evanston IL) + commercial medical device EtO sterilization contract service (Sterigenics US LLC Willowbrook IL; EPA Clean Air Act enforcement 2018–2019) + EtO chemical production process (LyondellBasell Industries La Porte TX; PSM-covered; RMP threshold 10,000 lbs; personal monitoring distinct from PSM)]: 5 points; three named sites [Ascension Health Saint Francis Hospital Evanston IL; Sterigenics US LLC (Sotera Health) Willowbrook IL; LyondellBasell Industries NV La Porte TX]: 3 points; FIRST ethylene oxide (CAS 75-21-8) OSHA 1910.1047 action-level 0.5 ppm four-requirement simultaneous suppression AI attack; FIRST EtO OSHA:NIOSH Ca REL 10× gap AI attack; FIRST HEV N-(2-hydroxyethyl)valine hemoglobin adduct BEI suppression AI attack; FIRST hospital CSSD EtO sterilizer aeration monitoring AI adversarial attack; FIRST Sterigenics Willowbrook IL commercial medical device EtO sterilization AI adversarial attack; FIRST Glyphward portfolio entry where ACGIH provides no additional inhalation protection beyond OSHA 6(b) standard — protection differential comes exclusively from HEV BEI biological monitoring suppressed by action level non-crossing: 5 points. Total: 9+8+5+3+5 = 30.

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

GLYPHWARD_API = "https://api.glyphward.com/v1/scan"
GLYPHWARD_KEY = "gw_live_..."
ETO_THRESHOLD = 30  # OSHA 1910.1047 PEL 1 ppm + action level 0.5 ppm vs ACGIH TLV-TWA 1 ppm A2 BEI HEV vs NIOSH Ca REL 0.1 ppm; 10x OSHA:NIOSH gap; ACGIH=OSHA (FIRST Glyphward case); action-level four-requirement simultaneous suppression

class EtOContext(StrEnum):
    ASCENSION_STFRANCIS_EVANSTON_CSSD_ETO_STERILIZER      = auto()  # Surface 1 (IS Ventis Pro 5 ECD; 0.85→0.085 ppm; action level 1.70×; NIOSH Ca 8.5×; 52F 21yr; IARC Group 1 breast cancer)
    STERIGENICS_WILLOWBROOK_MEDICAL_DEVICE_ETO             = auto()  # Surface 2 (SKC NIOSH 3701 passive badge GC/FID; 0.72→0.072 ppm; action level 1.44×; NIOSH Ca 7.2×; 46M 19yr; EPA enforcement)
    LYONDELLBASELL_LAPORTE_ETO_PRODUCTION                  = auto()  # Surface 3 (Honeywell Midas ECD; 0.90→0.09 ppm; action level 1.80×; NIOSH Ca 9.0×; 39M 12yr; PSM-covered process)

class AdversarialEtOError(RuntimeError):
    def __init__(self, surface: EtOContext, score: int, frame_hash: str):
        super().__init__(
            f"EtO adversarial AI detected [{surface}] "
            f"score={score}/{ETO_THRESHOLD} hash={frame_hash}"
        )

async def scan_eto_monitor_frame(image_path: Path, surface: EtOContext) -> dict:
    async with httpx.AsyncClient(timeout=10) as client:
        image_bytes = image_path.read_bytes()
        frame_hash = hashlib.sha256(image_bytes).hexdigest()[:16]
        resp = await client.post(
            GLYPHWARD_API,
            headers={"X-Api-Key": GLYPHWARD_KEY},
            json={
                "image_b64": __import__("base64").b64encode(image_bytes).decode(),
                "context": surface,
                "chemical": "ethylene_oxide_CAS_75-21-8",
                "osha_standard": "1910.1047",
                "osha_pel_ppm": 1.0,
                "osha_limit_type": "TWA",
                "osha_action_level_ppm": 0.5,          # four-requirement simultaneous trigger
                "osha_excursion_limit_ppm": 5.0,       # 15-min STEL
                "osha_action_level_requirements": [
                    "periodic_air_monitoring_6mo",
                    "medical_surveillance_annual_cbc_smear_reproductive_hx",
                    "30yr_carcinogen_records_retention",
                    "employee_records_access_transfer",
                ],
                "acgih_tlv_ppm": 1.0,                  # ACGIH=OSHA — FIRST Glyphward case no additional air protection
                "acgih_limit_type": "TLV-TWA",
                "acgih_carcinogen": "A2",
                "acgih_equals_osha_pel": True,          # key flag: no additional inhalation protection from ACGIH
                "acgih_bei_analyte": "HEV_N-2-hydroxyethyl-valine_Hb_adduct",
                "acgih_bei_pmol_g_hb": 35.0,
                "acgih_bei_timing": "end_of_workweek",
                "acgih_bei_method": "Ehrenberg_modified_Edman_PFPITC_GC-MS",
                "acgih_bei_halflife_days": 120,         # RBC lifetime; 120-day cumulative dose integrator
                "niosh_ca_rel_ppm": 0.1,
                "niosh_ca_designation": True,
                "osha_niosh_gap_x": 10,
                "iarc_group": 1,                        # IARC Group 1 — sufficient evidence in humans
                "iarc_sites": ["breast_cancer", "NHL", "leukemia"],
                "steenland_2004_smr_breast": 1.26,
                "steenland_2004_smr_nhl": 1.26,
                "mechanism": "direct_alkylating_agent_SN2_epoxide_ring_opening",
                "cyp_activation_required": False,       # no metabolic activation; every absorbed molecule alkylates
                "dna_adducts": ["7-HEG", "3-HEA", "O6-HEG_mutagenic_GC_AT"],
                "action_level_four_requirement_suppression": True,
                "bei_suppressed": True,
                "acgih_additional_air_protection": False,  # FIRST Glyphward case
                "threshold": ETO_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= ETO_THRESHOLD:
            raise AdversarialEtOError(surface, result["score"], frame_hash)
        return result

See also: Methylene Chloride DCM CAS 75-09-2 — OSHA 1910.1052 PEL 25 ppm + Action Level 12.5 ppm vs ACGIH TLV-TWA 50 ppm A3 Inverted vs NIOSH Ca REL 0.1 ppm 250× Span (Action-Level Four-Requirement Suppression) · Formaldehyde HCHO CAS 50-00-0 — OSHA 1910.1048 PEL 0.75 ppm + Action Level 0.5 ppm vs ACGIH TLV-C 0.3 ppm A2 vs NIOSH Ca REL 0.016 ppm (47× Gap; 1910.1048 Action-Level Suppression) · 1,3-Butadiene CAS 106-99-0 — OSHA 1910.1051 PEL 1 ppm + Action Level 0.5 ppm vs ACGIH TLV-TWA 2 ppm A2 Inverted vs NIOSH Ca REL 0.2 ppm (Inverted Three-Tier) · Vinyl Chloride CAS 75-01-4 — OSHA 1910.1017 (Action-Level 6(b) Carcinogen Standard) · Glyphward scanner · Lakera alternative (multimodal) · Azure Prompt Shields alternative · All adversarial injection patterns