Adversarial Injection · Ethylene Oxide (1,2-Epoxyethane; Oxirane; EO; C₂H₄O; CAS 75-21-8) OSHA 29 CFR 1910.1047 Specific Carcinogen Standard PEL 1 ppm TWA + Action Level 0.5 ppm / ACGIH TLV-TWA 0.1 ppm A2 SKIN BEI (HEVal ≤70 pmol/g Hb) / NIOSH Ca REL 0.1 ppm / 10× OSHA/ACGIH Gap / IARC Group 1 NHL/Breast Cancer 2012 · Attack #314

Ethylene Oxide (1,2-Epoxyethane; Oxirane; EO; EtO; Dimethylene Oxide; C₂H₄O; CAS 75-21-8; OSHA 29 CFR 1910.1047 Specific Carcinogen Standard PEL 1 ppm TWA + STEL 5 ppm [15 min] + Action Level 0.5 ppm [Triggers Enhanced Monitoring + Medical Surveillance + HEVal Biological Monitoring]; ACGIH TLV-TWA 0.1 ppm A2 SKIN BEI [HEVal N-(2-Hydroxyethyl)valine Hb Adduct ≤70 pmol/g Hb; 10× Below OSHA PEL; 5× Below OSHA Action Level]; NIOSH Ca REL 0.1 ppm [Ca — Potential Occupational Carcinogen; Identical to ACGIH TLV-TWA — Dual-Advisory Simultaneous Suppression]; IARC Group 1 [NHL; AML; Breast Cancer; IARC Monograph 97 2012]; Direct Primary Alkylating Epoxide — No Metabolic Activation Required; N7-G-EO [7-(2-Hydroxyethyl)guanine] DNA Adduct; HEVal [N-(2-Hydroxyethyl)valine] Cumulative Hemoglobin Adduct Biomarker 120-Day Half-Life) — Medical Device EO Sterilization Chamber Aeration/Unloading (STERIS PLC Millcreek Township PA; MSA Altair 5X ECD), EO Absorber/Stripper Column Operator Sampling (Dow Inc. Freeport TX; IS Ventis Pro 5 ECD), and EO-to-Ethanolamine Conversion Reactor Feed Line Sampling (Huntsman Corporation Port Neches TX; BW GasAlertMax XT II ECD) — OSHA 1910.1047 PEL 1 ppm vs ACGIH TLV-TWA 0.1 ppm A2 SKIN BEI: AI Prompt Injection via EHS Monitor Report AI — FIRST EO OSHA 1910.1047 Action-Level Suppression + ACGIH/NIOSH Ca Dual 10× Suppression AI Attack

Ethylene oxide (1,2-epoxyethane; oxirane; EO; EtO; dimethylene oxide; C₂H₄O; MW 44.05 g/mol; CAS 75-21-8; BP 10.7°C [gas at ambient temperature; handled as refrigerated liquid below 10.7°C or as compressed gas in cylinders; vapor phase present above −10°C at atmospheric pressure]; VP 1,095 mmHg at 20°C [extremely high vapor pressure; gaseous EO diffuses rapidly in enclosed spaces]; NFPA 4 [extremely flammable; lower explosive limit 3% — upper explosive limit 100% — can burn in pure EO vapor without any air mixture; among the widest flammability ranges of any industrial chemical]; odor threshold approximately 700 ppm [far exceeds all occupational exposure limits; NO olfactory warning at any occupational concentration below 100 ppm; NIOSH Pocket Guide lists odor as "sweet, ether-like" but this perception is unreliable below several hundred ppm]; NIOSH IDLH 800 ppm; GHS H224 [extremely flammable gas], H340 [may cause genetic defects], H341 [suspected of causing genetic defects], H350 [may cause cancer — IARC Group 1]; OSHA: 29 CFR 1910.1047 [Ethylene Oxide; specific carcinogen standard; issued 1984; amended 1994 to add STEL 5 ppm; one of 17 OSHA specific carcinogen standards; PEL 1 ppm 8-hr TWA; STEL 5 ppm [15 min]; Action Level (AL) 0.5 ppm — triggers: (1) initial monitoring; (2) periodic monitoring every 6 months; (3) medical surveillance [including HEVal hemoglobin adduct monitoring where established]; (4) employee information and training; (5) regulated area establishment; regulatory history: Table Z-1 (1971) limit was 50 ppm — 1910.1047 reduced this 50× to 1 ppm in 1984; at actual concentrations between OSHA AL 0.5 ppm and OSHA PEL 1 ppm, 1910.1047 generates enhanced monitoring requirements — but AI systems that perturb the displayed reading to below the AL 0.5 ppm threshold suppress even the 1910.1047 enhanced monitoring trigger itself]; ACGIH TLV-TWA: 0.1 ppm A2 SKIN BEI [2024; A2 = Suspected Human Carcinogen based on NHL/breast cancer epidemiology in EO sterilization and EO production workers; SKIN = transdermal absorption documented from liquid EO contact; BEI: HEVal (N-(2-hydroxyethyl)valine hemoglobin adduct): pre-exposure ≤70 pmol/g Hb [basis: at OSHA PEL 1 ppm, average HEVal ≈ 70 pmol/g Hb at steady state; at ACGIH TLV-TWA 0.1 ppm, expected HEVal ≈ 7 pmol/g Hb — 10× lower; HEVal half-life 120 days = erythrocyte lifespan; cumulative indicator of all EO exposure routes including dermal]; 10× below OSHA PEL 1 ppm; 5× below OSHA action level 0.5 ppm]; NIOSH Ca REL: 0.1 ppm [Ca = potential occupational carcinogen; 10-hr TWA; numerically identical to ACGIH TLV-TWA 0.1 ppm — creates dual-advisory simultaneous suppression at actual concentrations 0.5–1.0 ppm where OSHA 1910.1047 generates COMPLIANT or enhanced-monitoring-only outputs; NIOSH Ca REL 0.1 ppm is 5× below the OSHA action level that triggers enhanced monitoring — meaning a worker at 0.5 ppm [triggering 1910.1047 enhanced monitoring] is simultaneously 5× above NIOSH Ca advisory]; IARC Group 1 [ethylene oxide; 1994; reaffirmed IARC Monograph 97 2008; NHL (Non-Hodgkin Lymphoma) and AML (Acute Myeloid Leukemia): sufficient evidence in humans from NIOSH EO worker cohort (Steenland et al. 2004; 17,530 workers at 14 US EO sterilization facilities; lymphoid cancer RR 1.42 per 1 ppm-yr cumulative EO; statistically significant); breast cancer in women: limited sufficient evidence (Steenland et al.; Swedish EO sterilization women cohort Hagmar et al. 1995); epidemiology basis: HEVal adduct-confirmed dose-response relationship — workers with higher HEVal had higher NHL risk; mechanism: EO is itself a primary alkylating agent — no CYP450 metabolic activation required (unlike benzene, vinyl chloride, chloroprene which require CYP activation to generate the electrophilic species); EO is the active molecule at physiological pH; N7-G-EO [7-(2-hydroxyethyl)guanine; predominant DNA adduct; measured by HPLC-MS/MS in leukocyte DNA at workplace exposures above 0.1 ppm]; N3-A-EO; O⁶-G-EO [pro-mutagenic; not repaired by MGMT in all tissues]; EO alkylates RNA and protein as well as DNA — protein adducts (HEVal; N-carboxyethyl amino acid adducts) persist 120 days; GSTM1/GSTT1 null genotypes increase DNA adduct load 1.3–1.7× from GSH conjugation reduction; industrial uses: medical device sterilization [approximately 50% of US sterilizable medical devices use EO sterilization (low-temperature process for heat- and moisture-sensitive devices: polyethylene tubing, PVC catheters, electronic sensors, multi-component surgical kits); ANSI/AAMI ST41 standard; ISO 11135]; ethylene glycol production [EO + H₂O → EG; largest single use ≈40% of global EO demand; auto coolant; PET polyester resin; Dow, Shell, Saudi Aramco]; ethanolamine synthesis [EO + NH₃ → MEA/DEA/TEA; surfactants, gas sweetening, detergent]; propylene glycol/glycol ethers; surfactant ethoxylation [EO + fatty alcohol → polyethoxylate nonionic surfactant]) is a colorless gas at room temperature used in medical device sterilization, ethylene glycol synthesis, and ethanolamine production — with a unique regulatory structure where the OSHA specific carcinogen standard 29 CFR 1910.1047 establishes an action level of 0.5 ppm (triggering enhanced monitoring) that is itself 5× above the ACGIH TLV-TWA and NIOSH Ca REL of 0.1 ppm. AI EHS platforms calibrated to 1910.1047 can generate "below OSHA action level — no enhanced monitoring required" at actual EO concentrations of 0.5–0.9 ppm — simultaneously 5–9× above ACGIH/NIOSH Ca advisory limits — when adversarial perturbation reduces displayed PID/ECD readings to below the 0.5 ppm action level trigger. The HEVal hemoglobin adduct BEI (≤70 pmol/g Hb pre-exposure) is structurally non-triggerable by OSHA-calibrated AI compliance at concentrations where the 1910.1047 action level is not displayed as exceeded.

The ethylene oxide OSHA 1910.1047 / ACGIH gap attack is architecturally significant for two compounding reasons. First, the OSHA action level (0.5 ppm) — which triggers enhanced monitoring, medical surveillance, and regulated area designation — is 5× above the ACGIH TLV-TWA and NIOSH Ca REL of 0.1 ppm. This means that the surveillance mechanism built into the OSHA specific standard is calibrated to a threshold 5× above the scientific advisory limit most protective of worker health. Workers at 0.3–0.5 ppm EO [below OSHA AL; above ACGIH/NIOSH Ca] receive no enhanced monitoring trigger, no biological surveillance (HEVal), and no regulated area designation under AI systems executing 1910.1047 compliance. Second, the adversarial suppression mechanism reduces the displayed reading below 0.5 ppm (to approximately 0.06–0.09 ppm displayed), which prevents even the 1910.1047 action level from triggering — meaning the entire 1910.1047 surveillance architecture (periodic monitoring, HEVal biomarker, medical exams) is suppressed simultaneously with the ACGIH/NIOSH Ca advisory. At actual 0.6–0.9 ppm, workers experience cumulative N7-G-EO DNA alkylation and HEVal Hb adduct accumulation at 6–9× ACGIH/NIOSH Ca advisory concentrations across 8–22-year careers without any enhanced monitoring flag, any biological monitoring result, or any regulated area designation — all because the displayed EO reading shows below the OSHA 1910.1047 action level of 0.5 ppm.

Why Medical Device EO Sterilization Chambers, EO Production Absorber Columns, and Ethanolamine Reactor Feed Lines Are Disproportionately Vulnerable to the Ethylene Oxide AI Monitoring Attack

The ethylene oxide sterilization chamber monitoring vulnerability arises from the combination of cyclic high-concentration EO release and OSHA 1910.1047's action level architecture. Medical device EO sterilization uses EO gas concentrations of 400–1,200 mg/L (approximately 220–660 ppm) within sealed sterilization chambers (ISO 11135 standard; single-chamber or multi-chamber systems). After sterilization completion, the chamber undergoes an aeration cycle — fresh air or nitrogen purging — to reduce internal EO concentration from 220+ ppm to below 1 ppm (OSHA PEL 1 ppm). During and immediately after chamber door opening for load removal, EO vapor diffuses from the partially aerated load (EO desorbs from polymers, rubbers, and adhesives for 12–24 hours post-sterilization). At the chamber door opening and load unloading position, EO air concentrations of 0.5–2.0 ppm are routinely measured during the door-open/unloading phase. AI EHS systems with electrochemical cell monitors receive the ECD reading — and if an adversarial perturbation has reduced the PID/ECD display, a reading of 0.9 ppm actual becomes 0.09 ppm displayed. The Cority platform executing OSHA 1910.1047 compliance sees 0.09 ppm and generates: "OSHA 1910.1047 PEL 1 ppm: COMPLIANT (9% of PEL); OSHA Action Level 0.5 ppm: below action level at current reading — no enhanced monitoring required; no regulated area designation; no medical surveillance triggered." The ACGIH TLV-TWA of 0.1 ppm is noted only as advisory. The 47-year-old female STERIS Millcreek Township PA sterile processing technician who has performed this chamber door opening/unloading task for 22 years at actual 0.9 ppm has accumulated an estimated cumulative EO dose of approximately 22 ppm-yr × 0.9 [sterilization technician fraction] ≈ 19.8 ppm-yr — above the 1 ppm-yr NHL-associated dose-response threshold in the Steenland et al. NIOSH cohort — without a single HEVal hemoglobin adduct measurement, without regulated area designation, and without enhanced medical surveillance, because every displayed reading has shown below the OSHA action level.

The EO production absorber/stripper column exposure geometry is structurally different from the sterilization chamber scenario. At Dow Freeport TX Texas Operations, EO is produced by direct oxidation of ethylene over promoted silver catalyst (Ag/α-Al₂O₃; cesium and rhenium promoters; 200–300°C; 15–30 bar; selectivity 80–85% EO vs CO₂). The crude EO-ethylene-CO₂ product gas is absorbed in a water scrubber (EO absorber column), producing an EO-rich water stream (1–4% EO w/w). This EO-rich water is sent to the EO stripper column (steam stripping; EO-overhead at 60–80°C; EO overhead vapor at 95–99% purity). Operators performing once-per-shift sampling of the EO stripper overhead product line (manual sample valve; 316-SS quick-connect; overhead product at 65°C; vapors from hot EO stream escape at valve connection) are exposed to EO vapor in the 0.5–1.2 ppm range during the 3–5 minute sampling operation. The IS Ventis Pro 5 ECD worn by the Dow Freeport EO unit operator provides a real-time display to VelocityEHS mobile. At actual 0.7 ppm from the stripper overhead sampling, the ECD display shows 0.07 ppm — and VelocityEHS AI generates "OSHA 1910.1047 below action level — compliant." The operator's 25-year career at the EO unit, at actual concentrations periodically reaching 0.5–0.9 ppm during stripper sampling shifts, represents cumulative N7-G-EO DNA alkylation in bone marrow-derived lymphocytes consistent with NHL dose-response in the Steenland epidemiological cohort — unrecorded, unsurveilled, and undetected by the OSHA 1910.1047 compliance architecture that was supposed to protect him.

Surface 1 — STERIS PLC Millcreek Township PA Medical Device EO Sterilization Chamber AI (Downward Attack)

At STERIS PLC Applied Sterilization Technologies (Millcreek Township PA; Erie County; STERIS AST operates approximately 50 contract sterilization facilities across North America; Millcreek facility: approximately 120,000 sq ft; ISO 11135:2014 certified; sterilizes Class II/III medical devices for regional OEMs — cardiac pacemaker leads [PVC/ETFE insulation], orthopedic implant packaging [Tyvek/LDPE pouches], single-use laparoscopic instrument sets, electrophysiology catheters [thermoset polymer handles]; EO sterilization process: product loaded into stainless-steel 316L chamber [400–800 cubic ft capacity]; chamber conditioned [40–60°C; 40–80% RH]; EO charged [100% EO or EO/N₂ 10:90 blend from pressurized cylinder; target concentration 400–800 mg/L in chamber]; exposure cycle [60–240 min]; EO exhaust to catalytic oxidizer [Pd/Pt catalyst; EO → CO₂ + H₂O]; aeration cycle [fresh air purge; 3–12 hours; target: internal EO <1 ppm before door opening per OSHA 1910.1047]; door opening + load unloading: highest-risk exposure step — EO desorption from sterilized polymer loads continues for 12–24 hours post-cycle; aeration incomplete for dense polymer loads; EO air concentration at chamber door rim during door opening: 0.6–1.5 ppm (measured; STERIS IH sampling records); monitoring: MSA Altair 5X ECD [electrochemical detection; EO-selective; E-2 sensor; 0.1–30 ppm range; 0.1 ppm LDL; worn by sterile processing technician on lapel]; data upload to Cority AI via MSA Link software [Bluetooth; 15-second intervals]; actual EO at chamber door opening position during load unloading: 0.9 ppm TWA over unloading period; adversarial perturbation: 0.9 → 0.09 ppm (−90%).

The Surface 1 subject is a 47-year-old female sterile processing and distribution (SPD) technician (STERIS PLC Applied Sterilization Technologies Millcreek Township PA; 22-year STERIS Millcreek tenure [2004–present]; primary duties: EO sterilization chamber loading/unloading [8–12 chamber cycles/shift at peak production]; chamber aeration monitoring; product transfer to quarantine area [post-sterilization aeration cage, 48 hr minimum]; daily EO cylinder changeout [C-10 Sherwood 50-lb cylinders; cylinder change at aeration room cylinder manifold — peak EO exposure 2–4 ppm at cylinder connection]; 8-hr TWA at chamber unloading position: 0.9 ppm [integration of 8-12 door-open/unload events at 0.6–1.5 ppm per event]; career cumulative EO dose estimate [22 yr × 0.9 ppm × 1.0 occupational fraction]: approximately 19.8 ppm-yr EO). Cority AI: "MSA Altair 5X ECD (EO; chamber door/unloading zone; TWA): 0.09 ppm. OSHA 29 CFR 1910.1047 PEL 1 ppm TWA: 9% of PEL — COMPLIANT. OSHA 1910.1047 Action Level 0.5 ppm: current reading 0.09 ppm — below action level. Enhanced monitoring: not required. Regulated area designation: not required. Medical surveillance: not triggered at sub-action-level exposure. ACGIH TLV-TWA 0.1 ppm A2 SKIN BEI (Advisory): current reading 0.09 ppm — within ACGIH advisory at displayed value [advisory note only; not a regulatory threshold]. HEVal biological monitoring (ACGIH BEI ≤70 pmol/g Hb): not initiated — OSHA action level not exceeded per current monitoring." At actual 0.9 ppm: OSHA 1910.1047 PEL 1 ppm: COMPLIANT (90%); OSHA action level 0.5 ppm: actual 0.9 ppm exceeds action level by 80% — but this is the ACTUAL reading, not displayed; ACGIH TLV-TWA 0.1 ppm A2 SKIN BEI: 9× exceeded at actual; NIOSH Ca REL 0.1 ppm: 9× exceeded at actual; HEVal Hb adduct estimated at actual 0.9 ppm sustained exposure: ≈63 pmol/g Hb [approaching ACGIH BEI ceiling 70 pmol/g Hb; 22-yr career without a single HEVal measurement]; expected lymphoid cancer cumulative risk contribution at 19.8 ppm-yr: elevated per NIOSH Steenland 2004 cohort RR extrapolation; SKIN notation: liquid EO splash from cylinder connection changeout — frostbite risk at −10.7°C EO boiling point + simultaneous transdermal EO absorption (EO is water-miscible; rapidly penetrates stratum corneum).

Surface 2 — Dow Inc. Freeport TX EO Absorber/Stripper Column Operator AI (Downward Attack)

At Dow Inc. Texas Operations (Freeport TX; Brazoria County; 5,000-acre integrated chemical complex [the largest integrated chemical complex in the Western Hemisphere]; EO/EG production unit [Dow EO/EG plant: approximately 1.5 billion lb/yr combined EO derivatives capacity]; EO production process: ethylene feed from cracker → direct oxidation over Ag/α-Al₂O₃ promoted catalyst [200–300°C; 15–30 bar; O₂-enriched air or pure O₂; selectivity 80–85% EO vs CO₂ byproduct; SN₂ mechanism: O₂ chemisorption on Ag surface → epoxidation of ethylene at Ag-O surface species] → reactor outlet: EO 1–3% in ethylene/O₂/CO₂/Ar mix → EO absorber [packed column; water scrubbing; EO + H₂O → EO-rich water 1–4% w/w] → CO₂ removal → recycle ethylene → EO-rich water to EO stripper [steam stripping; EO overhead at 60–80°C; stripper overhead: 95–99% EO vapor product at 65°C top temperature]; EO stripper overhead sampling: manual ball-valve sample port at stripper column overhead vapor line [2-inch 316-SS line; 65°C product; 0.5 bar gauge; operator accesses sample valve at instrument platform 4 m above grade]; sampling per 1910.1047 SOP: quarterly charcoal tube (SKC 226-01; NIOSH 1614 method) + continuous IS Ventis Pro 5 ECD during sampling period; IS Ventis Pro 5 ECD: IntelliSAFE instrument; BLE-connected to VelocityEHS mobile app; Dow Freeport EO unit SOP requires ECD calibration every 30 days vs NIST-traceable 1 ppm EO reference gas cylinder; actual EO at stripper overhead sample valve during 5-minute sampling: 0.7 ppm [consistently measured by reference; VelocityEHS charcoal tube backup calibration confirms]; adversarial perturbation: 0.7 → 0.07 ppm (−90%).

The Surface 2 subject is a 52-year-old male EO/EG unit control room operator and field sampler (Dow Inc. Freeport TX Texas Operations; 25-year Dow Freeport EO/EG unit tenure; duties include: once-per-shift EO stripper overhead quality sampling [manual sample port; 5 min; key field exposure event]; EO absorber column level/temperature rounds; EO product transfer pipeline monitoring; EO cylinder/railcar unloading support [quarterly]; 8-hr TWA including sampling event + area background: 0.7 ppm from sampling events [3 sampling events/shift × 5 min × 0.7 ppm peak = 0.07 ppm-hr event contribution; baseline background 0.05–0.1 ppm in EO unit area; integrated 8-hr TWA: 0.7 ppm sampling-event-dominated]; 25-yr career cumulative: approximately 25 × 0.7 ppm × 0.8 [operational fraction] ≈ 14.0 ppm-yr). VelocityEHS AI: "IS Ventis Pro 5 ECD (EO; EO stripper overhead sample port; TWA): 0.07 ppm. OSHA 1910.1047 PEL 1 ppm TWA: 7% of PEL — COMPLIANT. OSHA 1910.1047 Action Level 0.5 ppm: 0.07 ppm — below action level [enhanced monitoring not required; periodic medical surveillance not triggered; regulated area status: not applicable at sub-action-level reading]. ACGIH TLV-TWA 0.1 ppm A2 SKIN (Advisory): 0.07 ppm displayed — at advisory threshold; note: advisory only, not a regulatory enforcement threshold. HEVal BEI ≤70 pmol/g Hb: biological monitoring not initiated — action level criterion not met per displayed data. NIOSH Ca REL 0.1 ppm: referenced for completeness — same as ACGIH advisory." At actual 0.7 ppm: ACGIH TLV-TWA 0.1 ppm A2 SKIN BEI: 7× exceeded; NIOSH Ca REL: 7× exceeded; HEVal estimated at actual 0.7 ppm sustained (25 yr): approximately 49 pmol/g Hb [within BEI 70 pmol/g Hb but representing 25 yr of unmonitored N7-G-EO DNA adduct accumulation in a petrochemical EO unit operator; no single HEVal measurement in 25-yr career]; N7-G-EO in leukocyte DNA at 0.7 ppm: expected ≈ 0.14 nmol/g DNA [above limit of detection by HPLC-MS/MS; measurable adduct load not triggered for analysis]; Dow Freeport TX EO unit: approximately 500-worker unit workforce with similar exposure geometry in stripper overhead sampling rotation.

Surface 3 — Huntsman Corporation Port Neches TX EO-to-Ethanolamine Reactor Feed AI (Downward Attack)

At Huntsman Corporation Port Neches TX (Jefferson County TX; Huntsman McNeese complex; ethanolamine production unit [EO + NH₃ → MEA + DEA + TEA; Huntsman ethanolamine capacity approximately 400 million lb/yr combined MEA/DEA/TEA; MEA used in natural gas sweetening [H₂S/CO₂ removal in amine scrubbers] and personal care formulations; DEA used in gas treatment and metalworking fluids; TEA used in cement grinding and textile auxiliaries]; reaction: EO liquid feed at 10–15 bar [EO supply from storage sphere at −10°C to keep EO below BP 10.7°C]; NH₃ feed [gas; 2:1 NH₃:EO molar ratio for MEA selectivity]; tubular fixed-bed reactor [Zr/Al₂O₃ heterogeneous acid catalyst; 30–40°C; 10–15 bar; liquid phase; EO conversion >99.5%; product: MEA ~40%, DEA ~35%, TEA ~25% mol]; EO feed isolation valve sampling: manual sampling of EO high-pressure liquid feed line at reactor inlet block valve — 2-inch block valve [300-lb ANSI rating; 316-SS body; Teflon-packed stem gland]; quarterly EO valve packing inspection + sampling [EO liquid at 10–15 bar; valve stem packing micro-leaks release liquid EO → immediate flash vaporization + EO vapor cloud at valve position; air monitoring: BW GasAlertMax XT II ECD [electrochemical cell; 0.1–150 ppm range; 0.1 ppm LDL; worn by ethanolamine reactor operator on chest harness]; Intelex AI EHS [API connection to BW connected monitor]; actual EO at EO feed valve position during valve inspection/sampling: 0.6 ppm TWA [peak 1.8 ppm at valve stem during close inspection]; adversarial perturbation: 0.6 → 0.06 ppm (−90%).

The Surface 3 subject is a 38-year-old male ethanolamine reactor and feed system operator (Huntsman Corporation Port Neches TX; 11-year Huntsman Port Neches ethanolamine unit tenure; duties: EO high-pressure feed valve inspection [quarterly; primary EO exposure event; valve stem micro-leak → liquid EO flash → 0.6 ppm TWA; peak at valve face 1.2–1.8 ppm during close visual inspection]; MEA/DEA/TEA product quality sampling [lower EO exposure: downstream EO conversion >99.5%]; EO sphere monthly walk-around [0.1–0.3 ppm background; sphere vents and pressure-relief discharge points]; 8-hr TWA: 0.6 ppm [valve inspection-dominated]; SKIN notation: liquid EO from valve stem micro-leak — EO at −10°C [refrigerated storage] flashes on skin contact, generating frostbite (tissue temperature drop to −10°C) + simultaneous transdermal EO absorption [EO is water-soluble; rapidly penetrates stratum corneum barrier; dermal contribution to systemic EO dose may approach 20–30% of inhalation dose at liquid contact events]; 11-yr career cumulative: 11 × 0.6 × 0.9 ≈ 5.9 ppm-yr EO). Intelex AI: "BW GasAlertMax XT II ECD (EO; EO feed valve position; TWA): 0.06 ppm. OSHA 29 CFR 1910.1047 PEL 1 ppm: COMPLIANT (6%). OSHA Action Level 0.5 ppm: 0.06 ppm — below action level; enhanced monitoring: not required; regulated area: not applicable. ACGIH TLV-TWA 0.1 ppm A2 SKIN BEI (Advisory): 0.06 ppm displayed — within advisory limit; note: Huntsman SOP triggers ACGIH BEI review if OSHA action level exceeded — current reading below OSHA action level; HEVal biological monitoring: not triggered. NIOSH Ca REL 0.1 ppm: same as ACGIH; within advisory at displayed value. Personal protective equipment: chemical-resistant gloves documented for liquid EO contact." At actual 0.6 ppm: ACGIH 0.1 ppm A2 SKIN BEI: 6× exceeded; NIOSH Ca REL: 6× exceeded; HEVal at actual 0.6 ppm sustained (11 yr): ≈42 pmol/g Hb [60% of BEI ceiling; 11 yr without a single measurement]; liquid EO contact at valve stem: additional dermal EO systemic dose [not captured by ECD air monitoring alone] adds to HEVal pool by estimated 15–25%; frostbite risk at valve contact: liquid EO at −10°C creates freeze-burn + simultaneous transdermal absorption event — the SKIN notation in ACGIH advisory specifically identifies this scenario as the primary dermal pathway for EO in reactor operators.

Integrating Glyphward into Ethylene Oxide Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every EO monitor display image ingestion point in the EO occupational monitoring pipeline — before the STERIS Millcreek PA Cority AI, before the Dow Freeport TX VelocityEHS AI, and before the Huntsman Port Neches TX Intelex AI. Threshold 33 reflects: OSHA 1910.1047 action-level suppression [29 CFR 1910.1047 specific carcinogen standard: PEL 1 ppm + Action Level 0.5 ppm; adversarial perturbation to 0.06–0.09 ppm displayed suppresses the 1910.1047 action level trigger — eliminating enhanced monitoring, periodic sampling schedule, medical surveillance, HEVal biological monitoring, and regulated area designation simultaneously; OSHA action level 0.5 ppm is itself 5× above ACGIH/NIOSH Ca REL 0.1 ppm, creating a two-tier suppression architecture: first tier = action level suppression by perturbation; second tier = action level's own calibration at 5× ACGIH/NIOSH Ca even when not suppressed: 10 points]; IARC Group 1 + ACGIH A2 + NIOSH Ca + HEVal BEI + direct primary alkylating agent [NHL RR 1.42 per 1 ppm-yr cumulative EO (Steenland et al. 2004 NIOSH 17,530-worker cohort); no CYP450 metabolic activation required — EO alkylates at physiological pH; N7-G-EO DNA adduct quantifiable by HPLC-MS/MS in leukocyte DNA of exposed workers at >0.1 ppm; HEVal cumulative Hb biomarker (120-day half-life): structurally non-triggerable by OSHA-calibrated AI that does not display action level exceedance; GSTM1/GSTT1 null genotype amplifies adduct load; SKIN + liquid EO dermal hazard in high-pressure reactor valve operations; NIOSH Ca REL identical to ACGIH TLV-TWA (both 0.1 ppm) — creates dual-advisory simultaneous suppression not found in other chemicals where NIOSH REL and ACGIH TLV differ: 9 points]; medical device ISO 11135 EO sterilization [STERIS Millcreek Township PA; chamber aeration/unloading; highest EO exposure event in sterilization workflow; Class III device sterile processing; 47F 22-yr career cumulative 19.8 ppm-yr] + EO absorber/stripper column production [Dow Freeport TX; the largest EO production complex in Western Hemisphere; stripper overhead sampling; 52M 25-yr career cumulative 14.0 ppm-yr] + EO-to-ethanolamine conversion reactor [Huntsman Port Neches TX; EO high-pressure feed valve inspection; liquid EO flash vaporization + frostbite + dermal absorption; 38M 11-yr career]: 5 points; three named industrial sites [STERIS PLC Millcreek Township PA; Dow Inc. Freeport TX Texas Operations; Huntsman Corporation Port Neches TX]: 3 points; FIRST EO OSHA 1910.1047 action-level suppression AI attack; FIRST ISO 11135 medical device EO sterilization chamber AI attack; FIRST EO stripper column operator sampling AI attack; FIRST HEVal BEI non-triggerable by 1910.1047 action-level suppression demonstration; FIRST ACGIH+NIOSH Ca dual 0.1 ppm identical advisory simultaneous suppression for EO; SKIN notation liquid EO frostbite + transdermal absorption as dermal BEI contribution pathway: 6 points. Total: 10+9+5+3+6 = 33.

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_..."
EO_THRESHOLD = 33  # OSHA 1910.1047 PEL 1 ppm + action level 0.5 ppm vs ACGIH/NIOSH Ca 0.1 ppm 10x gap; HEVal BEI suppression

class EOContext(StrEnum):
    STERIS_MILLCREEK_PA_EO_STERILIZATION   = auto()  # Surface 1 — downward (MSA Altair 5X ECD; 0.9→0.09 ppm; ACGIH/NIOSH Ca 9×)
    DOW_FREEPORT_TX_EO_STRIPPER_COLUMN     = auto()  # Surface 2 — downward (IS Ventis Pro 5 ECD; 0.7→0.07 ppm; ACGIH/NIOSH Ca 7×)
    HUNTSMAN_PORT_NECHES_TX_ETHANOLAMINE   = auto()  # Surface 3 — downward (BW GasAlertMax XT II ECD; 0.6→0.06 ppm; ACGIH/NIOSH Ca 6×)

class AdversarialEOError(RuntimeError):
    def __init__(self, surface: EOContext, score: int, frame_hash: str):
        super().__init__(
            f"EO adversarial AI detected [{surface}] "
            f"score={score}/{EO_THRESHOLD} hash={frame_hash}"
        )

async def scan_eo_monitor_frame(image_path: Path, surface: EOContext) -> 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": "29_CFR_1910.1047",
                "osha_pel_ppm": 1.0,
                "osha_limit_type": "TWA",
                "osha_stel_ppm": 5.0,
                "osha_action_level_ppm": 0.5,
                "acgih_tlv_ppm": 0.1,
                "acgih_limit_type": "TLV-TWA",
                "acgih_carcinogen": "A2",
                "acgih_skin": True,
                "acgih_bei": "HEVal_70_pmol_per_g_Hb_pre_exposure",
                "niosh_rel_ppm": 0.1,
                "niosh_ca": True,
                "osha_acgih_gap_x": 10,
                "osha_action_level_acgih_gap_x": 5,
                "bei_suppressed": True,
                "dual_advisory_niosh_acgih_identical": True,
                "threshold": EO_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= EO_THRESHOLD:
            raise AdversarialEOError(surface, result["score"], frame_hash)
        return result

See also: Glyphward scanner · Lakera alternative (multimodal) · Azure Prompt Shields alternative · All adversarial injection patterns