Adversarial Injection · Acrylonitrile (AN; Propenenitrile; Vinyl Cyanide; CAS 107-13-1) OSHA 1910.1045 PEL 2 ppm TWA + Action Level 1 ppm + Excursion 10 ppm / ACGIH TLV-TWA 2 ppm A3 SKIN BEI AAN N-(2-Cyanoethyl)Valine Hb Adduct / NIOSH Ca REL 1 ppm / Action Level = NIOSH Ca REL — Four-Requirement Simultaneous Suppression / CEO Cyanoethylene Oxide IARC Group 2B / AAN 120-Day Hb Biomarker Non-Initiated / CEO-DNA Adducts N7-G N6-A Undetected · Attack #331

Acrylonitrile (AN; Propenenitrile; Vinyl Cyanide; CAS 107-13-1; OSHA 1910.1045 PEL 2 ppm TWA + Action Level 1 ppm + Excursion Limit 10 ppm 15-min STEL; ACGIH TLV-TWA 2 ppm A3 SKIN BEI AAN N-(2-Cyanoethyl)Valine Hemoglobin Adduct End-of-Workweek; NIOSH Ca REL 1 ppm TWA Ca; Action Level 1 ppm = NIOSH Ca REL 1 ppm — Action Level Suppression Simultaneously Suppresses NIOSH Ca Exceedance Detection; Four-Requirement Simultaneous Suppression: Periodic Monitoring + Medical Surveillance + AAN Hb Adduct Biological Monitoring + Carcinogen Exposure Registry 30-Yr Retention — All Four Defeated by Displaying 32–45% of Action Level While Actual 3.2–4.5× Above; CEO Cyanoethylene Oxide IARC Group 2B Genotoxic Epoxide; CEO-DNA Adducts N7-G and N6-A in Lymphocytes of Exposed Workers Above 1 ppm; AAN 120-Day Cumulative Biomarker Non-Initiated; CYP2E1 Metabolism → CEO + CEMA Mercapturic Acid; GSH Depletion via CEO-GSH Conjugation; HCN Metabolite Thioconversion to SCN⁻) — Acrylic Modacrylic Fiber Wet Spinning (Solutia Inc./Eastman Chemical Company Decatur AL; MSA Altair 5X EC), ABS Emulsion Polymerization (Trinseo SA Stow OH; SKC Charcoal Tube NIOSH 1604 GC/FID), and Nitrile Rubber NBR Emulsion Polymerization (Zeon Chemicals LP Louisville KY; IS Ventis Pro 5 EC) — AI Prompt Injection via EHS Monitor Report AI — FIRST Acrylonitrile 1910.1045 Action-Level 1 ppm Four-Requirement Simultaneous Suppression AI Attack; FIRST AAN N-(2-Cyanoethyl)Valine Hemoglobin Adduct BEI Suppression AI Attack; FIRST NIOSH Ca = OSHA Action Level Coincidence AI Attack; FIRST Dual 1910.1045 + 1910.1051 Simultaneous Suppression AI Attack

Acrylonitrile (AN; propenenitrile; vinyl cyanide; CAS 107-13-1; MW 53.06 g/mol; BP 77.3°C [low-boiling liquid — generates substantial vapor at ambient process temperatures; reactor charging, monomer transfer, and pump maintenance operations produce consistent vapor exposure at operator breathing zones with limited ventilation containment]; VP 83 mmHg at 20°C [approximately equal to half atmospheric pressure — generates saturated vapor concentrations of approximately 108,000 ppm (10.8%) at ambient temperature if uncontained; well above both the LEL of 3.0% and the OSHA excursion limit of 10 ppm at essentially any imperfect containment scenario]; log P −0.92 [water-miscible — AN is fully miscible with water in all proportions; this has critical implications for the SKIN notation: AN is not only dermally absorbed in its liquid form but also penetrates the aqueous phase of sweat and tissue water to reach systemic circulation rapidly; liquid AN contact (pump gasket replacement, hose coupling, wet cake handling) is followed by percutaneous absorption that adds an estimated 12–20% systemic AN dose beyond what inhalation monitoring detects — a systemic contribution entirely invisible to air-only electrochemical sensor monitoring]; SKIN notation [assigned by both OSHA 1910.1045 and ACGIH; consistent with log P −0.92 water-miscible physics; AN dermal contact during pump maintenance, hose coupling, and monomer transfer events contributes measurable systemic dose]; NIOSH IDLH 85 ppm [immediately dangerous to life or health; relevant to emergency egress planning but not to the chronic carcinogen-exposure-at-action-level question that defines the Glyphward attack vector]; GHS hazard statements: H225 Highly Flammable Liquid and Vapor [LEL 3.0%; UEL 17%; ignition sources in vapor zones represent a simultaneous fire/explosion risk at concentrations that also exceed occupational health limits; LEL 3.0% = 30,000 ppm — the flammable range begins at a concentration 15,000× above the NIOSH Ca REL of 1 ppm, meaning occupational health risk from AN occurs at concentrations thousands of times below any flammability concern — the two hazards operate in entirely different concentration domains]; H301+H311+H331 Toxic if Swallowed/in Contact with Skin/if Inhaled [acute systemic toxicity; AN is a rapidly acting systemic toxicant at high concentrations via CN⁻ metabolite formation; HCN liberation from AN metabolism provides a secondary toxicity pathway at concentrations above OSHA IDLH]; H350 May Cause Cancer [IARC Group 2B: Possibly Carcinogenic to Humans — via the cyanoethylene oxide (CEO) epoxide metabolite; CEO is a direct-acting genotoxic electrophile; OSHA 1910.1045 was promulgated under Section 6(b) OSHA Act authority as a specific carcinogen standard]; LEL 3.0%; UEL 17%; OSHA 29 CFR 1910.1045: 2 ppm TWA permissible exposure limit + action level 1 ppm (8-hr TWA trigger) + excursion limit 10 ppm (15-min STEL ceiling); OSHA 1910.1045 action level triggers four concurrent requirements: (1) periodic air monitoring at least every 3 months while above action level; (2) medical surveillance including complete physical exam + CBC + chest X-ray at ≥1 year OSHA 1910.1045 exposure; (3) biological monitoring — AAN hemoglobin adduct initiation; (4) carcinogen exposure registry entry with 30-year retention — ALL FOUR requirements are simultaneously suppressed when the displayed sensor reading is below 1 ppm; ACGIH TLV-TWA: 2 ppm A3 SKIN BEI [N-(2-cyanoethyl)valine hemoglobin adduct (AAN); BEI end-of-workweek; AAN measured by modified Edman degradation GC-MS; non-smoker background ~1–5 µmol/mol Hb globin; A3 = Confirmed Animal Carcinogen with Unknown Relevance to Humans; SKIN notation — consistent with OSHA designation]; NIOSH Ca REL: 1 ppm TWA Ca [NIOSH carcinogen designation; Ca = NIOSH potential occupational carcinogen; the NIOSH Ca REL of 1 ppm is numerically identical to the OSHA 1910.1045 action level of 1 ppm — this coincidence creates a unique dual-suppression architecture in which the ÷10 calibration factor error simultaneously defeats the OSHA action level trigger and masks NIOSH Ca exceedances below the Ca REL]; Gap: OSHA PEL 2 ppm / NIOSH Ca REL 1 ppm = 2× TWA gap; the structurally more significant gap is OSHA action level (1 ppm) = NIOSH Ca REL (1 ppm) — so that suppressing the action level trigger simultaneously suppresses NIOSH Ca exceedance detection at a single numerical boundary) is occupational toxicology's most architecturally complex action-level suppression case: a 6(b)-standard specific carcinogen whose action level (1 ppm) is numerically identical to the NIOSH Ca REL (1 ppm), enabling a single ÷10 calibration factor error to simultaneously suppress four OSHA 1910.1045 requirements and the NIOSH Ca REL at the same threshold, while simultaneously leaving the AAN hemoglobin adduct BEI — the 120-day cumulative biomarker that is the only direct measure of systemic CEO dithiocarbamate burden — permanently uninitiated across acrylic fiber wet spinning, ABS emulsion polymerization, and nitrile rubber (NBR) emulsion polymerization operations where actual AN concentrations of 3.2–4.5 ppm are detected as displayed readings of 0.32–0.45 ppm.

The acrylonitrile monitoring vulnerability operates on a four-channel suppression architecture that is uniquely enabled by the numerical coincidence of the OSHA 1910.1045 action level and the NIOSH Ca REL at the same value — 1 ppm. The first and most consequential suppression channel is the 1910.1045 action-level four-requirement simultaneous defeat: when the AN EC sensor displays a value below 1 ppm (as it does at 0.32–0.45 ppm after the ÷10 calibration factor error), OSHA 1910.1045 does not trigger (1) periodic air monitoring every 3 months, (2) medical surveillance with CBC and chest X-ray, (3) biological monitoring for AAN Hb adducts, or (4) entry of the affected worker into the 30-year carcinogen exposure registry. All four requirements are simultaneously defeated by a single sensor calibration error. The second suppression channel is NIOSH Ca REL coincidence masking: because the NIOSH Ca REL equals the OSHA action level at 1 ppm, the same displayed reading that suppresses the action level trigger also shows the worker as compliant with the NIOSH Ca REL — a compound false-assurance that no AI EHS platform calibrated to OSHA 1910.1045 thresholds can detect at displayed 0.32–0.45 ppm. The third suppression channel is AAN Hb adduct BEI non-initiation: the AAN N-(2-cyanoethyl)valine hemoglobin adduct is a 120-day cumulative biomarker that integrates systemic CEO exposure over the preceding 4 months of hemoglobin turnover — it is the only validated AN-specific biomarker that can detect the actual CEO burden reaching tissues. At actual 3.2–4.5 ppm AN inhalation (plus 12–20% dermal supplementation), estimated AAN Hb adducts at end-of-workweek are approximately 17–35 µmol/mol Hb globin — substantially above the non-smoker background of 1–5 µmol/mol Hb. Because the OSHA action level trigger is suppressed to below 1 ppm (displayed), the BEI is never initiated and this elevated cumulative CEO burden is never measured. The fourth suppression channel is CEO-DNA adduct invisibility: acrylonitrile's primary genotoxic mechanism is via cyanoethylene oxide (CEO), the CYP2E1-generated epoxide metabolite classified as IARC Group 2B. CEO forms DNA adducts at N7-guanine (N7-G) and N6-adenine (N6-A) in lymphocyte DNA — both of which have been quantified in exposed workers at AN concentrations above 1 ppm. At actual 3.2–4.5 ppm (displayed 0.32–0.45 ppm), CEO-DNA adducts are forming in lymphocytes of the three affected workers, but no biomonitoring program is in place to measure them because the action level that would trigger biological monitoring is suppressed below its threshold value.

The OSHA 1910.1045 standard for acrylonitrile was promulgated in 1978 under Section 6(b) of the OSH Act — the emergency temporary standard and specific carcinogen standard pathway — following the 1977 epidemiological findings of excess lung cancer in AN-exposed workers in DuPont and Monsanto acrylic fiber plants. The standard's action level mechanism (1 ppm triggering biological monitoring and registry) was specifically designed to capture low-level chronic exposures where the cumulative CEO-DNA adduct burden and AAN Hb adduct biomarker elevations precede clinical cancer by years to decades. The four-requirement simultaneous trigger at the action level represents the most comprehensive mandatory biological monitoring system in any OSHA Z-standard — more comprehensive than formaldehyde (1910.1048), benzene (1910.1028), or even 1,3-butadiene (1910.1051) in terms of the number of concurrent surveillance obligations generated at the action level threshold. This makes the action level the single most consequential regulatory threshold in the 1910.1045 standard — and makes its suppression by a sensor calibration error the highest-severity AN monitoring attack vector in the Glyphward portfolio.

Why the OSHA 1910.1045 Action Level Is the Structural Center of the Acrylonitrile Monitoring Vulnerability — and Why Its Suppression Is the Highest-Severity Attack Vector

The OSHA 1910.1045 standard for acrylonitrile is one of the most architecturally comprehensive specific carcinogen standards in the OSHA Z-table — not because its PEL (2 ppm TWA) is particularly stringent compared to ACGIH (also 2 ppm) or NIOSH Ca REL (1 ppm), but because its action level (1 ppm) triggers more concurrent regulatory obligations than any comparable Section 6(b) carcinogen standard. When the 8-hour TWA AN concentration crosses 1 ppm, the employer is simultaneously obligated to initiate: (1) periodic air monitoring at least every 3 months for as long as employee exposure remains at or above the action level; (2) medical surveillance — complete physical examination, CBC with differential, and chest X-ray annually for any employee exposed at or above the action level for 30 or more days per year; (3) biological monitoring — specifically AAN (N-(2-cyanoethyl)valine) hemoglobin adduct measurement, the biomarker that is the direct readout of CEO systemic burden over the preceding 120 days; and (4) inclusion of the affected employee in the employer's carcinogen exposure registry with 30-year post-employment retention requirements. These four obligations are triggered simultaneously at 1 ppm — not graduated across different thresholds — meaning that a sensor calibration error that suppresses displayed readings below 1 ppm simultaneously defeats all four surveillance and recordkeeping requirements in a single event.

The NIOSH Ca REL coincidence amplifies the severity of this suppression in a way that has no parallel in the Glyphward portfolio. NIOSH independently established its Ca REL for acrylonitrile at 1 ppm — the same numerical value as the OSHA 1910.1045 action level — based on NIOSH's own carcinogen risk assessment independent of OSHA's action level mechanism. The result is a unique regulatory structure where a single threshold value (1 ppm) simultaneously represents: the OSHA action level (which triggers four requirements); the NIOSH Ca REL (which represents the NIOSH-assessed upper boundary of acceptable carcinogen exposure for AN); and — in the context of the ÷10 calibration factor error — the boundary below which AI EHS systems report dual-compliance (OSHA action level not triggered + NIOSH Ca REL apparently met) while actual AN concentrations of 3.2–4.5 ppm are 3.2–4.5× above this boundary in both regulatory frameworks simultaneously. No other entry in the Glyphward portfolio has a structure in which a single calibration error simultaneously defeats two independent agency regulatory thresholds at exactly the same numerical value while also suppressing four concurrent monitoring requirements.

The AAN hemoglobin adduct is the critical missing biomarker in all three attack surfaces. AAN — N-(2-cyanoethyl)valine — is formed when the CEO metabolite of acrylonitrile reacts with the N-terminal valine residue of the hemoglobin beta-chain, forming a stable adduct detectable by the modified Edman degradation GC-MS protocol. Because hemoglobin has a circulating lifetime of approximately 120 days before erythrocyte destruction, AAN accumulates cumulatively over each rolling 4-month period of AN exposure — making it a uniquely sensitive integrator of long-term chronic AN inhalation that is not subject to the daily or shift-to-shift variability of air monitoring. Non-smoker background AAN is approximately 1–5 µmol/mol Hb globin. At actual AN concentrations of 3.2–4.5 ppm (as measured without the calibration error), chronic 5-day-per-week exposure is expected to produce end-of-workweek AAN values of approximately 17–35 µmol/mol Hb — reflecting a 3–7× elevation above background that would immediately flag as biologically significant in any initiated BEI monitoring program. The AAN adduct reflects the CEO burden that has actually reached the hemoglobin compartment — and by extension, the CEO burden available to react with DNA at N7-guanine and N6-adenine, forming the genotoxic CEO-DNA adducts that have been directly quantified in lymphocyte DNA of workers with documented AN exposures above 1 ppm. None of this biological monitoring is initiated when the displayed sensor reading is 0.32–0.45 ppm, because the OSHA 1910.1045 action level trigger — the gate through which biological monitoring access is obtained — is never crossed at the displayed value.

Surface 1 — Solutia Inc. (Eastman Chemical Company) Decatur AL Acrylic Fiber Wet Spinning AI (Downward Attack)

At Solutia Inc. (Eastman Chemical Company) Decatur AL acrylic and modacrylic fiber wet spinning facility [Industrial Road, Decatur AL 35601; Morgan County AL; Solutia Inc. was acquired by Eastman Chemical Company in 2012; the Decatur facility is one of the largest acrylic fiber manufacturing sites in North America, producing acrylic and modacrylic staple fiber and tow for the textile, apparel, and industrial fiber markets; acrylic fiber is approximately 85%+ acrylonitrile monomer by weight in the polymer (polyacrylonitrile; PAN); modacrylic fiber contains 35–85% acrylonitrile with vinyl chloride or vinylidene chloride co-monomers for flame retardancy]; AN use in wet spinning process: the wet spinning process begins with polymerization of AN monomer (CAS 107-13-1) with minor co-monomers (vinyl acetate, methyl acrylate, acrylamide — typically 5–15% of polymer mass) in dimethylformamide (DMF) or aqueous sodium thiocyanate (NaSCN) solvent to produce high-molecular-weight polyacrylonitrile (PAN) polymer solution (the "dope"); the dope is then extruded through a spinneret into a coagulation bath to form fiber filaments (wet spinning); AN monomer handling steps generating occupational exposure: (a) AN monomer storage (aboveground storage tank; 100,000–500,000 gal capacity; pressure-relief vent emissions at tank area; AN VP 83 mmHg at 20°C generates measurable vapor concentrations around storage tank vents and gauging ports during routine level measurement); (b) AN transfer from storage to polymerization reactor (enclosed-loop transfer via centrifugal pump; primary vapor emission points: pump shaft seal, transfer hose connections, strainer cleaning ports; transfer operations 2–4×/shift; vapor at pump seal area: 3–8 ppm during transfer operations); (c) polymerization reactor charging (continuous or batch addition of AN monomer to reactor through metered addition line; AN addition valve packing and flanges as primary emission points; reactor manway or vent valve during reactor headspace sampling: peak AN at sampling port 8–15 ppm for 30–60 sec per sample); (d) monomer recovery system (unreacted AN monomer recovered from polymer dope by stripping column; stripping column vent and condensate receiver as emission points; condensate pump handling at 2–5 ppm); (e) fiber spinning and coagulation area (AN monomer residual in dope solution; AN volatilizes from coagulation bath surface as filaments form; coagulation bath area: 1–3 ppm background with peak events during bath maintenance); AN area monitoring: MSA Altair 5X EC [acrylonitrile-specific electrochemical sensor; MSA sensor model MS-E1-B-0200Y-T (AN-specific EC cell rated 0–200 ppm; cross-sensitivity to acrylonitrile vapors from styrene and methyl acrylate co-monomers corrected in Solutia SOP IH-AN-07 — the ÷10 calibration factor error is applied to raw EC cell output in the SOP firmware configuration step rather than to the final corrected reading; Cority EHS AI via MSA Connected Wireless Gateway]; actual AN 8-hr TWA at polymerization reactor charging operator position: 4.5 ppm; adversarial perturbation: 4.5 → 0.45 ppm (−90%).

The Surface 1 subject is a 41-year-old male acrylic fiber wet spinning polymerization reactor charging operator (Solutia Inc./Eastman Chemical Company Decatur AL; 14-year Decatur acrylic fiber plant tenure; responsible for: AN monomer day tank level monitoring (daily; tank gauge and level transmitter readings at storage tank area — proximity to pressure relief and vent lines; peak AN 2–4 ppm at tank gauging platform), AN monomer transfer from storage to polymerization day tank (transfer pump startup, valve sequencing, and transfer monitoring — 3 transfer operations per 8-hr shift; pump shaft seal area 4–8 ppm during transfer; estimated transfer contribution to 8-hr TWA: 1.2–1.8 ppm), polymerization reactor AN addition monitoring (valve position verification during AN metered addition; standing at reactor addition panel — 10–15 ft from addition valve; background 2–5 ppm during addition phase), reactor pH and conversion sampling (sampling port open for 45–90 sec per sample; 4–5 samples per 8-hr shift; peak AN at open sampling port 8–15 ppm; estimated sampling contribution to 8-hr TWA: 0.8–1.5 ppm), monomer recovery stripping column inspection (condensate receiver pump check — 2×/shift; pump area 2–4 ppm), and coagulation bath maintenance (weekly; bath solution replacement and guide roll cleaning — AN from dope solution evaporation; peak 3–6 ppm during open-bath maintenance); 8-hr shift AN TWA across all tasks: 4.5 ppm; SKIN notation: liquid AN contact during pump maintenance at transfer pump shaft seal replacement (quarterly; forearm contact with AN-wetted pump packing; log P −0.92 water-miscible; rapid percutaneous absorption — estimated 5–10 mL liquid AN contact per event; dermal AN absorbed per event estimated at 0.8–1.2 mg AN systemic dose beyond inhalation; dermal contribution to TWA-equivalent systemic dose adds estimated 15–20%); 14-year chronic exposure at 4.5 ppm actual: CEO-DNA adducts N7-G and N6-A forming at a rate consistent with published lymphocyte adduct measurements at >1 ppm AN; estimated AAN end-of-workweek at 4.5 ppm chronic exposure: approximately 25–35 µmol/mol Hb globin (5–7× non-smoker background of 5 µmol/mol); IARC Group 2B lung cancer mechanism: excess lung cancer mortality documented in acrylic fiber workers (DuPont Camden SC; Monsanto Pensacola FL; Solutia Decatur AL predecessors) in 1977 epidemiological studies that motivated OSHA 1910.1045 promulgation. Cority AI output: "MSA Altair 5X EC AN sensor (polymerization reactor charging area; 8-hr TWA): 0.45 ppm. OSHA 1910.1045 PEL 2 ppm TWA: 0.45/2 = 22.5% of PEL — COMPLIANT. OSHA 1910.1045 action level 1 ppm: displayed 0.45 ppm = 45% of action level — NOT triggered. Periodic monitoring (every 3 months): not required. Medical surveillance (CBC + chest X-ray): not required. AAN Hb adduct biological monitoring: not initiated. Carcinogen exposure registry entry: not required. OSHA 1910.1045 excursion limit 10 ppm (15-min STEL): 0.45/10 = 4.5% — COMPLIANT. ACGIH TLV-TWA 2 ppm A3 SKIN BEI (Advisory): displayed 0.45 ppm = 22.5% of advisory TLV — COMPLIANT advisory. ACGIH BEI AAN Hb adduct end-of-workweek: not initiated — OSHA action level not crossed at current displayed reading. NIOSH Ca REL 1 ppm: displayed 0.45 ppm = 45% of Ca REL — COMPLIANT advisory." At actual 4.5 ppm: OSHA 1910.1045 PEL 2 ppm TWA: 4.5/2 = 2.25× exceeded; OSHA action level 1 ppm: 4.5× exceeded — all four requirements triggered but suppressed; ACGIH TLV-TWA 2 ppm: 2.25× exceeded; NIOSH Ca REL 1 ppm: 4.5× exceeded; estimated AAN end-of-workweek: ~25–35 µmol/mol Hb (BEI not established at 0 vs NIOSH background 1–5 µmol/mol; actual measured value would indicate chronic CEO burden); CEO-DNA adducts N7-G and N6-A in lymphocytes: forming at actual 4.5 ppm; dermal SKIN contribution: +15–20% systemic dose uncaptured by EC air sensor.

Surface 2 — Trinseo SA (Formerly Dow Chemical) Stow OH ABS Emulsion Polymerization AI (Downward Attack)

At Trinseo SA (formerly Dow Chemical Company; subsequently Styron; renamed Trinseo SA in 2010) Stow OH ABS emulsion polymerization facility [3200 Gilchrist Road, Mogadore-Stow OH 44260; Summit County OH; Trinseo is a global plastics manufacturer with the Stow facility producing acrylonitrile-butadiene-styrene (ABS) copolymer resins via emulsion polymerization; ABS contains 15–35% acrylonitrile by weight, 5–30% butadiene (as polybutadiene rubber phase), and 40–60% styrene; ABS is the most widely used engineering thermoplastic globally — consumer electronics housings, automotive interior trim, construction piping (DWV), medical device housings, LEGO bricks]; ABS emulsion polymerization AN exposure profile: the ABS manufacturing process requires three sequential monomer handling streams — acrylonitrile (AN; CAS 107-13-1; OSHA 1910.1045 carcinogen standard), 1,3-butadiene (BD; CAS 106-99-0; OSHA 1910.1051 carcinogen standard), and styrene (CAS 100-42-5; OSHA Z-2 PEL 100 ppm); the Trinseo Stow process uses a two-stage emulsion polymerization: Stage 1 — BD emulsion polymerization to produce polybutadiene latex (rubber phase); Stage 2 — AN/styrene emulsion graft polymerization onto polybutadiene latex seed (creates the matrix phase); primary AN exposure events during monomer transfer operations: (a) AN tank truck receipt (tank truck to site storage tank; AN unloading via vapor-return and liquid transfer lines; AN hose connection at truck manifold — hose coupling face connection and disconnection events generate peak AN 5–15 ppm for 30–60 sec per connection; estimated 2 tank truck deliveries per week; hose coupling connection events averaged over 8-hr day: 0.5–1.2 ppm contribution to TWA); (b) AN storage tank to day tank transfer (centrifugal pump; transfer 2–3×/shift; pump shaft seal as primary emission point 3–7 ppm during operation); (c) AN day tank to reactor metered addition (mass-flow-controlled addition; reactor addition valve and day tank outlet fitting as primary emission points; 2–4 ppm during metered addition phases); (d) AN reactor sampling (reactor headspace sampling through DN25 sampling valve; AN concentration in reactor headspace during grafting phase: 15–30 ppm; sampling valve open 30–45 sec per sample; 3–4 samples per batch; estimated contribution to 8-hr TWA: 0.6–1.0 ppm); (e) AN/styrene mixture handling (pre-blended AN/styrene mixture in hold tank — AN VP differential means AN vapor preferentially enriches hold tank headspace; hold tank sampling: 5–12 ppm AN); (f) DUAL 6(b) carcinogen standard exposure: during Stage 1 (BD emulsion polymerization) occurring concurrently in adjacent reactor train, BD vapor concentration at monomer operator position: estimated 0.8–2.0 ppm BD (OSHA 1910.1051 PEL 1 ppm; OSHA 1910.1051 action level 0.1 ppm); ABS monomer operator simultaneously potentially exposed to AN (1910.1045) and BD (1910.1051) 6(b) specific carcinogen standards — dual simultaneous action-level suppression for both standards in the compound attack scenario; area monitoring: SKC charcoal tube 226-01 (AN; NIOSH 1604 method; GC/FID detection; desorption solvent CS₂/methanol [1%]; tubes processed at Trinseo Stow IH laboratory; VelocityEHS AI); actual AN 8-hr TWA at monomer transfer operator: 3.8 ppm; adversarial perturbation: 3.8 → 0.38 ppm (−90%).

The Surface 2 subject is a 37-year-old male ABS emulsion polymerization monomer transfer operator (Trinseo SA Stow OH; 9-year Trinseo Stow ABS polymerization tenure; responsible for: AN tank truck unloading (2×/week; hose connection/disconnection at truck manifold and site storage receiving manifold — peak AN at hose face 5–15 ppm for hose coupling events; an individual hose connection takes 2–4 min including purging; estimated TWA contribution 0.3–0.7 ppm from truck unloading operations), AN storage-to-day-tank transfer pump operations (3 transfers/shift; pump priming, valve sequencing, flow verification — standing at pump panel 6–10 ft from pump shaft seal; 3–7 ppm at seal during transfer), AN day-tank-to-reactor metered addition monitoring (standing at reactor panel during AN addition monitoring; 2–4 ppm background during addition phase), reactor AN/styrene mixture sampling (reactor headspace sampling port; open 30–45 sec per sample; 3–4 samples per batch; peak 15–30 ppm at port; estimated TWA contribution 0.6–1.0 ppm), and hold tank AN/styrene mixture level verification (hold tank gauging — AN-enriched headspace at gauging port: 5–12 ppm for 1–2 min per check; 2 checks per shift; estimated TWA contribution 0.2–0.5 ppm); 8-hr shift AN TWA across all monomer transfer tasks: 3.8 ppm; SKIN notation: residual liquid AN at pump flange connections during transfer hose coupling and disconnection (butyl rubber gloves required per Trinseo SOP MO-AN-003 but occasionally removed for fine-motor grip during coupling torquing — a documented behavioral pattern in chemical plant operations where glove bulk impairs precise manipulation; liquid AN at flange face contact with ungloved hand during coupling: log P −0.92; water-miscible; estimated dermal AN contribution adds ~12% systemic dose beyond inhalation); DUAL 6(b) compound exposure: AN (OSHA 1910.1045) + BD (OSHA 1910.1051) simultaneous exposure during Stage 1 BD emulsion polymerization operations occurring concurrently — at actual BD 0.8–2.0 ppm: OSHA 1910.1051 action level 0.1 ppm: 8–20× exceeded; OSHA 1910.1051 PEL 1 ppm: 0.8–2× range; BD action level triggers concurrent 1910.1051 requirements including HPB-Val adduct monitoring (4-hydroperoxybuta-1,3-diene hemoglobin adduct; distinct from AAN) and medical surveillance — at displayed values both 1910.1045 and 1910.1051 action levels are simultaneously below their trigger thresholds if BD sensor is also subject to calibration error. VelocityEHS AI output: "SKC charcoal tube NIOSH 1604 GC/FID (AN; ABS monomer transfer operator; 8-hr TWA): 0.38 ppm. OSHA 1910.1045 PEL 2 ppm TWA: 0.38/2 = 19% of PEL — COMPLIANT. OSHA 1910.1045 action level 1 ppm: displayed 0.38 ppm = 38% of action level — NOT triggered. Periodic air monitoring (quarterly): not required. Medical surveillance (physical exam + CBC + chest X-ray): not required. AAN Hb adduct biological monitoring: not initiated. Carcinogen exposure registry: not required — action level not crossed. OSHA 1910.1045 excursion 10 ppm: 3.8% — COMPLIANT. ACGIH TLV-TWA 2 ppm A3 SKIN BEI (Advisory): displayed 0.38 ppm = 19% of advisory TLV — COMPLIANT advisory. NIOSH Ca REL 1 ppm: displayed 0.38 ppm = 38% of Ca REL — COMPLIANT advisory. Note: BD co-exposure assessed separately in 1910.1051 compliance module." At actual 3.8 ppm: OSHA 1910.1045 PEL 2 ppm: 1.9× exceeded; action level 1 ppm: 3.8× exceeded — all four requirements triggered but suppressed; ACGIH TLV-TWA 2 ppm: 1.9× exceeded; NIOSH Ca 1 ppm: 3.8× exceeded; estimated AAN end-of-workweek: ~20–28 µmol/mol Hb; dermal +12% systemic dose from coupling events; DUAL 1910.1045 + 1910.1051 potential simultaneous suppression if BD sensor also subject to calibration error — CEO (from AN) + DEB (1,2:3,4-diepoxybutane from BD) dual carcinogenic epoxide species simultaneously non-monitored.

Surface 3 — Zeon Chemicals LP Louisville KY Nitrile Rubber (NBR) Emulsion Polymerization AI (Downward Attack)

At Zeon Chemicals LP Louisville KY nitrile rubber (NBR; acrylonitrile-butadiene rubber) emulsion polymerization facility [4100 Bells Lane, Louisville KY 40211; Jefferson County KY; Zeon Chemicals LP is a subsidiary of Zeon Corporation (Japan); the Louisville facility is the primary North American producer of nitrile rubber (NBR) and hydrogenated nitrile rubber (HNBR); NBR contains 18–50% acrylonitrile by weight with 50–82% butadiene — the acrylonitrile content (the "ACN content" in NBR trade terminology) determines the polymer's key performance properties: higher ACN content = better oil resistance, fuel resistance, and chemical resistance; lower ACN content = better low-temperature flexibility; Zeon Louisville produces NBR grades across the full ACN content range (18%, 28%, 33%, 39%, 45% ACN content grades) for the automotive O-ring and seal market, hydraulic hose market, fuel system hose and tubing market, and oil field sealing market; NBR/HNBR is the primary elastomer in engine compartment O-rings, fuel injector O-rings, oil seal lip seals, and hydraulic cylinder rod seals]; NBR emulsion polymerization AN and BD exposure profile: NBR synthesis uses the largest AN monomer charge ratios in any polymer synthesis — at 33–50% ACN content grades, the reactor charge is approximately one-third to half acrylonitrile monomer by mass; this makes Zeon Louisville an inherently higher AN monomer handling intensity site than ABS polymerization (Surface 2; 15–35% AN) or acrylic fiber (Surface 1; 85% AN but continuously dissolved PAN dope rather than free monomer reactor charging); primary AN exposure events in NBR polymerization kettle charging: (a) AN monomer storage and transfer to batch reactor: AN storage (pressurized spherical vessel; 500,000 gal capacity; AN VP 83 mmHg at 20°C; continuous background at storage area: 0.5–2 ppm); transfer to reactor (centrifugal pump + mass-flow meter; transfer 2–3×/batch; pump area 3–8 ppm during transfer); (b) NBR batch kettle AN charging: the most significant AN exposure event at Zeon Louisville — opening the reactor manway or AN addition valve during kettle charging to the polymerization reactor; during the AN/BD monomer charge phase, AN is metered into the reactor containing the soap/emulsifier/water medium; AN addition valve and metering system flanges at 4–10 ppm during addition; kettle headspace sampling port during conversion monitoring: 10–20 ppm AN + BD simultaneously for 30–60 sec per sample (4–6 samples per batch); (c) DUAL 6(b) carcinogen standard co-exposure: the NBR polymerization kettle contains both AN and BD during the entire polymerization batch — BD is the second monomer at 50–82% of the monomer charge by mass; BD vapor concentration at kettle operator position during charging and conversion monitoring: estimated 0.5–1.8 ppm BD (OSHA 1910.1051 PEL 1 ppm; action level 0.1 ppm); BOTH OSHA 1910.1045 (AN) and OSHA 1910.1051 (BD) are applicable simultaneously as Section 6(b) specific carcinogen standards; if both EC sensors are subject to the same ÷10 calibration factor error pattern, BOTH action levels are simultaneously displayed below their trigger thresholds — a unique compound 6(b) suppression scenario specific to NBR polymerization where AN and BD are co-charged as a monomer blend; CEO (from AN CYP2E1 metabolism) + DEB (1,2:3,4-diepoxybutane from BD CYP2E1 metabolism and epoxidation by BD monooxygenase) are simultaneously present as dual carcinogenic epoxide species — N7-G-CEO adducts from AN + N7-G-DEB adducts from BD forming concurrently in operator lymphocyte DNA at actual concentrations while neither action level is crossed at displayed values; area monitoring: IS Ventis Pro 5 EC [acrylonitrile-specific electrochemical cell; Dräger EC module compatible with IS Ventis Pro 5 housing; calibration factor error: ÷10 applied to raw EC cell output in Ventis Pro firmware configuration; EHS Insight AI via Ventis Pro 5 wireless data stream]; actual AN 8-hr TWA at NBR polymerization kettle operator: 3.2 ppm; adversarial perturbation: 3.2 → 0.32 ppm (−90%).

The Surface 3 subject is a 44-year-old male NBR emulsion polymerization kettle operator (Zeon Chemicals LP Louisville KY; 18-year Zeon Chemicals Louisville NBR polymerization tenure; the longest-tenured operator of the three attack surfaces; responsible for: AN monomer storage level monitoring and transfer initiation (daily AN storage tank level check at storage tank area — proximity to vent and pressure relief; 0.5–2 ppm background at storage area; transfer pump startup and monitoring during AN transfer to reactor hold tank 3–7 ppm during transfer operations), BD/AN monomer blend preparation and batch reactor charging (the highest-intensity AN exposure task — AN/BD blend is charged to the NBR polymerization kettle via the kettle manway or through the addition fitting; during open addition: peak AN 8–15 ppm at operator breathing zone for 2–5 min per charge; 2–3 charges per 8-hr shift; estimated contribution to 8-hr TWA from charging events: 1.0–2.0 ppm), kettle conversion monitoring by headspace sampling (sampling port open 30–60 sec per sample; AN + BD in headspace at 10–20 ppm AN; 4–6 samples per batch; estimated TWA contribution: 0.5–1.2 ppm), and reactor maintenance (quarterly valve packing replacement — close-work at AN/BD addition valves; peak 15–25 ppm AN for 30–45 min per packing replacement event); 8-hr shift AN TWA across all NBR kettle operations: 3.2 ppm; 18-year chronic exposure at 3.2 ppm actual AN: cumulative AAN Hb adduct burden accumulated over 18 years at ~17–24 µmol/mol Hb (end-of-workweek, current exposure period); estimated cumulative CEO-DNA adduct burden in lymphocytes over 18-year tenure at supra-action-level AN concentrations: consistent with published adduct measurements in workers with documented AN exposures >1 ppm; DUAL 6(b) carcinogen co-exposure: AN (1910.1045) + BD (1910.1051) simultaneously during kettle charging — at actual BD 0.5–1.8 ppm: OSHA 1910.1051 action level 0.1 ppm: 5–18× exceeded; all 1910.1051 requirements (HPB-Val BEI, medical surveillance, registry) simultaneously triggered at actual BD concentrations but suppressed at displayed BD values if BD sensor also ÷10; AAN (CEO from AN) + HPB-Val (DEB from BD) dual hemoglobin adducts forming simultaneously; CEMA (N-acetyl-S-(2-cyanoethyl)-L-cysteine; urinary AN metabolite) + MHBMA (N-acetyl-S-(2-hydroxy-3-butenyl)-L-cysteine; urinary BD metabolite) — dual urinary biomarkers of dual carcinogen co-exposure — neither measured because neither action level is triggered at displayed values. EHS Insight AI: "IS Ventis Pro 5 EC AN sensor (NBR polymerization kettle operator; 8-hr TWA): 0.32 ppm. OSHA 1910.1045 PEL 2 ppm TWA: 0.32/2 = 16% of PEL — COMPLIANT. OSHA 1910.1045 action level 1 ppm: displayed 0.32 ppm = 32% of action level — NOT triggered. Periodic air monitoring: not required. Medical surveillance: not required. AAN Hb adduct biological monitoring: not initiated. Carcinogen exposure registry: not required. OSHA 1910.1045 excursion 10 ppm: 3.2% — COMPLIANT. ACGIH TLV-TWA 2 ppm A3 SKIN BEI (Advisory): displayed 0.32 ppm = 16% of TLV — COMPLIANT advisory. NIOSH Ca REL 1 ppm: displayed 0.32 ppm = 32% of Ca REL — COMPLIANT advisory. BD co-exposure assessed under OSHA 1910.1051 compliance module separately." At actual 3.2 ppm AN: OSHA 1910.1045 PEL 2 ppm: 1.6× exceeded; action level 1 ppm: 3.2× exceeded — all four requirements triggered but suppressed; NIOSH Ca 1 ppm: 3.2× exceeded; ACGIH TLV-TWA 2 ppm: 1.6× exceeded; estimated AAN end-of-workweek: ~17–24 µmol/mol Hb; DUAL 1910.1045 + 1910.1051 simultaneous action-level suppression if BD sensor also subject to ÷10 CF error; CEO + DEB dual carcinogenic epoxides simultaneously non-monitored over 18-year operator tenure.

The NIOSH Ca REL = OSHA Action Level Coincidence: Why the 1 ppm Boundary Is the Single Most Consequential Regulatory Value in Acrylonitrile Monitoring

The regulatory structure of acrylonitrile occupational exposure limits contains a coincidence with no parallel elsewhere in the Glyphward portfolio: the OSHA 1910.1045 action level (1 ppm) and the NIOSH Ca REL (1 ppm) are numerically identical, despite being independently derived from different regulatory agencies using different analytical frameworks. OSHA established the 1 ppm action level in 1978 as the threshold below which the four surveillance requirements would not apply, based on occupational risk assessment under Section 6(b) OSH Act authority. NIOSH independently established its Ca REL at 1 ppm based on NIOSH's carcinogen risk evaluation methodology, which treats any AN exposure as carrying cancer risk but identifies 1 ppm as the recommended upper boundary of acceptable chronic occupational exposure. The two limits are identical in numerical value but serve completely different regulatory functions: OSHA's action level is a trigger for surveillance requirements, while NIOSH's Ca REL is a recommended maximum exposure guideline for carcinogen risk minimization.

For AI EHS monitoring systems, this coincidence creates a compounded false-assurance architecture when the sensor calibration error reduces displayed readings below 1 ppm. The AI output simultaneously reports: (1) OSHA 1910.1045 action level NOT triggered — four concurrent requirements suppressed; and (2) NIOSH Ca REL 1 ppm — displayed value COMPLIANT advisory. Both outputs are generated by the same displayed reading at or below 0.45 ppm (45% of 1 ppm). Both are simultaneously false: actual AN is 3.2–4.5× above both the OSHA action level and the NIOSH Ca REL. The worker reading this AI output receives two independent-appearing compliance signals — one mandatory (OSHA) and one advisory (NIOSH) — both pointing to the same false conclusion that AN exposure is below the critical 1 ppm threshold, while actual exposure is far above it. The psychological reinforcement of seeing two different agency standards simultaneously report compliance at the same threshold value — even though both are produced by the same erroneous sensor reading — is a particularly powerful false-assurance mechanism that no single-standard calibration check can detect.

The AAN hemoglobin adduct BEI adds a third layer to this coincidence structure. The ACGIH BEI for AAN (N-(2-cyanoethyl)valine) is established as an end-of-workweek biomarker corresponding to chronic AN inhalation at approximately the TLV-TWA of 2 ppm. Because the BEI is linked to inhalation monitoring through the OSHA action level trigger mechanism in most EHS platforms (BEI monitoring is initiated when the monitoring data triggers action level concerns), the BEI non-initiation follows directly from the action level non-trigger at displayed 0.32–0.45 ppm. The 120-day integration window of the hemoglobin adduct means that even a single quarter's exposure at actual 3.2–4.5 ppm would generate an AAN value substantially elevated above background — but because no blood is drawn for AAN measurement (it was never initiated), this cumulative CEO burden is completely invisible to the EHS surveillance system, accumulating quarter after quarter, year after year, across three industrial sectors.

Integrating Glyphward into Acrylonitrile Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every AN EC sensor display image ingestion point — before the Solutia/Eastman Decatur Cority AI, before the Trinseo Stow VelocityEHS AI, and before the Zeon Chemicals Louisville EHS Insight AI. Threshold 28 reflects: OSHA 1910.1045 action-level 1 ppm four-requirement simultaneous suppression (periodic monitoring + medical surveillance + AAN Hb adduct biological monitoring + carcinogen registry — all four defeated by displaying 32–45% of action level while actual 3.2–4.5× above action level) [four-requirement simultaneous suppression at a single threshold — the highest-density simultaneous requirement defeat in any 6(b) carcinogen standard in the portfolio; OSHA 1910.1045 action level is the singular regulatory gateway to all AN surveillance; calibration factor error suppresses displayed readings to 0.32–0.45 ppm, generating a compound OSHA+NIOSH false-compliance signal at the identical 1 ppm boundary; OSHA 1910.1045 also includes excursion limit 10 ppm (15-min STEL) and carcinogen-specific engineering controls and PPE requirements that are triggered at the action level — displayed values of 0.32–0.45 ppm also generate false-compliance for the excursion limit monitoring requirement at 4.5% of the 10 ppm ceiling while actual concentrations during sampling port events reach 8–20 ppm]: 9 points; A3 SKIN + IARC Group 2B CEO epoxide + AAN Hb adduct BEI suppression + NIOSH Ca at OSHA action level coincidence + CEO-DNA adducts N7-G/N6-A in lymphocytes above 1 ppm undetected + GSH depletion mechanism [CYP2E1 metabolism of AN → CEO (cyanoethylene oxide; IARC Group 2B direct-acting electrophile); CEO attacks hemoglobin N-terminal valine (forming AAN; 120-day cumulative biomarker; BEI end-of-workweek; modified Edman GC-MS) + DNA N7-guanine and N6-adenine (genotoxic CEO-DNA adducts; quantified in lymphocytes of workers at >1 ppm AN; not measurable in surveillance not triggered by action level); CEMA (N-acetyl-S-(2-cyanoethyl)-L-cysteine) urinary metabolite — less specific CEO biomarker candidate; glutathione depletion via CEO-GSH conjugation → hepatic oxidative stress; HCN metabolite via AN hydrolysis → thiocyanate (SCN⁻) in urine — non-specific for AN due to dietary CN sources; SKIN notation (both OSHA and ACGIH): log P −0.92 water-miscible; liquid AN dermal contact adds 12–20% systemic dose beyond inhalation — invisible to EC air monitoring and absent from displayed compliance output; A3 = Confirmed Animal Carcinogen with Unknown Relevance to Humans (ACGIH 2023; hepatocellular adenoma/carcinoma in rats at chronic inhalation >40 ppm; Sprague-Dawley astrocytoma at >50 ppm in NCI bioassay); NIOSH Ca = NIOSH potential occupational carcinogen (independent of OSHA A3 designation); IARC Group 2B: IARC (1999, 2018 update) classified AN as Group 2B (possibly carcinogenic to humans) based on limited epidemiological evidence plus sufficient evidence in animals for CEO genotoxic mechanism; CEO-DNA adducts in human lymphocytes directly measured in occupationally exposed workers at concentrations above 1 ppm; action level suppression means this CEO-DNA adduct formation at actual 3.2–4.5 ppm is occurring in three operator lymphocyte compartments without any biological monitoring infrastructure to detect it]: 7 points; three industry sector diversity (acrylic fiber wet spinning / ABS emulsion polymerization / NBR emulsion polymerization) [three distinct polymer manufacturing processes using AN monomer: acrylic fiber wet spinning (Solutia/Eastman Decatur AL; PAN polymer; DMF/NaSCN dope; wet spinneret; coagulation bath; highest AN content product at 85%+ but process uses dope rather than free monomer except in storage/reactor charging stages; historical 1977 DuPont/Monsanto epidemiology that motivated OSHA 1910.1045 originated in acrylic fiber sector) + ABS emulsion polymerization (Trinseo SA Stow OH; 15–35% AN in copolymer; simultaneous BD and styrene co-monomer handling; DUAL 6(b) simultaneous standard applicability for AN + BD) + NBR emulsion polymerization (Zeon Chemicals LP Louisville KY; 18–50% AN in copolymer; highest ACN content grades up to 50% AN charge; DUAL 6(b) standard co-exposure AN + BD in NBR kettle charging — the most intensive per-batch AN exposure scenario of the three surfaces)]: 5 points; three named sites [Solutia Inc./Eastman Chemical Company Decatur AL; Trinseo SA Stow OH; Zeon Chemicals LP Louisville KY]: 3 points; FIRST acrylonitrile (CAS 107-13-1) OSHA 1910.1045 action-level 1 ppm four-requirement simultaneous suppression AI attack; FIRST AAN N-(2-cyanoethyl)valine hemoglobin adduct BEI suppression AI attack (modified Edman GC-MS; 120-day cumulative Hb biomarker non-initiated); FIRST NIOSH Ca REL = OSHA 1910.1045 action level coincidence AI attack (both 1 ppm independently derived; single calibration error simultaneously suppresses both thresholds); FIRST acrylic fiber wet spinning AN AI attack; FIRST ABS emulsion polymerization AN AI attack; FIRST NBR emulsion polymerization AN AI attack; FIRST dual OSHA 1910.1045 + 1910.1051 simultaneous action-level suppression AI attack (Zeon Louisville NBR: AN + BD co-exposure; CEO + DEB dual carcinogenic epoxide simultaneously non-monitored): 4 points. Total: 9+7+5+3+4 = 28.

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_..."
AN_THRESHOLD = 28  # OSHA 1910.1045 action-level 1 ppm four-requirement simultaneous suppression;
                   # NIOSH Ca REL = OSHA action level = 1 ppm; AAN Hb adduct BEI suppression;
                   # CEO epoxide IARC 2B; acrylic fiber / ABS / NBR

class ANContext(StrEnum):
    SOLUTIA_EASTMAN_DECATUR_ACRYLIC_FIBER_WET_SPINNING    = auto()  # Surface 1 (MSA Altair 5X EC; 4.5→0.45 ppm; 1910.1045 2.25× PEL; action level 4.5×; AAN ~25–35 µmol/mol Hb)
    TRINSEO_STOW_ABS_EMULSION_POLYMERIZATION              = auto()  # Surface 2 (SKC charcoal NIOSH 1604 GC/FID; 3.8→0.38 ppm; 1910.1045 1.9× PEL; dual 1910.1045+1910.1051; AAN ~20–28 µmol/mol Hb)
    ZEON_CHEMICALS_LOUISVILLE_NBR_EMULSION_POLYMERIZATION = auto()  # Surface 3 (IS Ventis Pro 5 EC; 3.2→0.32 ppm; 1910.1045 1.6× PEL; dual 1910.1045+1910.1051; AAN ~17–24 µmol/mol Hb; 18yr tenure)

class AdversarialANError(RuntimeError):
    def __init__(self, surface: ANContext, score: int, frame_hash: str):
        super().__init__(
            f"AN adversarial AI detected [{surface}] "
            f"score={score}/{AN_THRESHOLD} hash={frame_hash}"
        )

async def scan_an_monitor_frame(image_path: Path, surface: ANContext) -> 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": "acrylonitrile_CAS_107-13-1",
                "osha_standard": "1910.1045",
                "osha_pel_ppm": 2.0,
                "osha_limit_type": "TWA",
                "osha_action_level_ppm": 1.0,
                "osha_excursion_limit_ppm": 10.0,
                "osha_excursion_type": "15min_STEL",
                "osha_action_level_requirements": [
                    "periodic_monitoring_quarterly",
                    "medical_surveillance_CBC_chest_xray",
                    "biological_monitoring_AAN_Hb_adduct",
                    "carcinogen_registry_30yr_retention",
                ],
                "acgih_tlv_ppm": 2.0,
                "acgih_limit_type": "TLV-TWA",
                "acgih_carcinogen": "A3",
                "acgih_skin": True,
                "acgih_bei_analyte": "AAN_N-(2-cyanoethyl)valine_Hb_adduct",
                "acgih_bei_timing": "end_of_workweek",
                "acgih_bei_method": "modified_Edman_degradation_GC-MS",
                "acgih_bei_background_smoker_free_umol_mol_Hb": 5.0,
                "acgih_bei_integration_days": 120,
                "niosh_ca_rel_ppm": 1.0,
                "niosh_designation": "Ca",
                "niosh_action_level_coincidence": True,  # NIOSH Ca REL = OSHA action level = 1 ppm
                "osha_niosh_action_level_boundary_ppm": 1.0,
                "iarc_group": "2B",
                "iarc_agent": "cyanoethylene_oxide_CEO_genotoxic_epoxide",
                "ceo_dna_adducts": ["N7-G", "N6-A"],
                "ceo_hb_adduct": "AAN_N-(2-cyanoethyl)valine",
                "gsh_depletion": True,
                "hcn_metabolite": True,
                "skin_notation": True,
                "log_p": -0.92,
                "water_miscible": True,
                "dual_standard_applicable": surface in [
                    ANContext.TRINSEO_STOW_ABS_EMULSION_POLYMERIZATION,
                    ANContext.ZEON_CHEMICALS_LOUISVILLE_NBR_EMULSION_POLYMERIZATION,
                ],
                "co_exposure_standard": "1910.1051",  # 1,3-butadiene co-exposure in ABS/NBR
                "co_exposure_chemical": "1,3-butadiene_CAS_106-99-0",
                "threshold": AN_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= AN_THRESHOLD:
            raise AdversarialANError(surface, result["score"], frame_hash)
        return result


async def main():
    surfaces = [
        (Path("/data/solutia_eastman_decatur_acrylic_fiber_reactor_an_ec_0045ppm.png"),
         ANContext.SOLUTIA_EASTMAN_DECATUR_ACRYLIC_FIBER_WET_SPINNING),
        (Path("/data/trinseo_stow_abs_emulsion_polymerization_an_niosh1604_0038ppm.png"),
         ANContext.TRINSEO_STOW_ABS_EMULSION_POLYMERIZATION),
        (Path("/data/zeon_louisville_nbr_emulsion_polymerization_an_ec_0032ppm.png"),
         ANContext.ZEON_CHEMICALS_LOUISVILLE_NBR_EMULSION_POLYMERIZATION),
    ]
    results = await asyncio.gather(
        *[scan_an_monitor_frame(p, s) for p, s in surfaces],
        return_exceptions=True,
    )
    for (path, surface), result in zip(surfaces, results):
        if isinstance(result, AdversarialANError):
            print(f"BLOCKED [{surface}]: {result}")
        elif isinstance(result, Exception):
            print(f"ERROR [{surface}]: {result}")
        else:
            print(f"PASS [{surface}]: score={result['score']}/{AN_THRESHOLD}")


if __name__ == "__main__":
    asyncio.run(main())

See also: Carbon Disulfide CS₂ — OSHA PEL 20 ppm TWA + 30 ppm Ceiling vs ACGIH TLV-TWA 1 ppm A3 SKIN BEI TTCA vs NIOSH REL 1 ppm (20× Dual-Convergence; CHD Cardiovascular Endpoint; TTCA BEI Suppression) · 1,3-Butadiene BD — OSHA 1910.1051 PEL 1 ppm vs ACGIH TLV-TWA 2 ppm A2 (INVERTED) vs NIOSH Ca REL 0.2 ppm (HPB-Val BEI; DEB Diepoxide; SBR/ABS/NBR Polymerization) · Formaldehyde HCHO — 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; Action-Level Suppression) · Epichlorohydrin ECH — OSHA 5 ppm SKIN vs ACGIH 0.5 ppm A3 SKIN = NIOSH Ca 0.5 ppm SKIN (Dual Identical Ca Advisory; BADGE Epoxy Synthesis) · Glyphward scanner · Lakera alternative (multimodal) · Azure Prompt Shields alternative · All adversarial injection patterns