Adversarial Injection · 1,2-Dichloroethane (EDC; Ethylene Dichloride; CAS 107-06-2) OSHA PEL 50 ppm TWA = NIOSH IDLH 50 ppm (Identical-Value PEL/IDLH Convergence) / ACGIH TLV-TWA 10 ppm A3 SKIN / NIOSH Ca REL 1 ppm / 50× OSHA:NIOSH Span / Chloroacetaldehyde CYP2E1 εA/εC Etheno-Adduct Suppression · Attack #320

1,2-Dichloroethane (EDC; Ethylene Dichloride; C₂H₄Cl₂; CAS 107-06-2; OSHA PEL 50 ppm TWA = NIOSH IDLH 50 ppm [Most Extreme PEL/IDLH Identical-Value Convergence in Z-1 Table]; ACGIH TLV-TWA 10 ppm A3 SKIN; NIOSH Ca REL 1 ppm; 50× OSHA:NIOSH Span; Chloroacetaldehyde εA/εC Etheno-Adduct; IARC Group 2A Cholangiocarcinoma) — Polycarbonate Interfacial Synthesis EDC Recovery Column (Covestro LLC Baytown TX; IS Ventis Pro 5 PID), EDC Railcar Storage and Transfer (Olin Corporation Freeport TX; SKC Charcoal GC/FID), and PVC Pipe Cement Manufacturing (Oatey Company Cleveland OH; MSA Altair 5X PID) — OSHA 50 ppm TWA vs ACGIH TLV-TWA 10 ppm A3 vs NIOSH Ca REL 1 ppm: AI Prompt Injection via EHS Monitor Report AI — FIRST EDC OSHA PEL = NIOSH IDLH Identical-Value Architectural Paradox + FIRST 50× EDC OSHA:NIOSH Ca Span + FIRST Polycarbonate Interfacial Synthesis EDC Solvent AI Attack

1,2-Dichloroethane (EDC; ethylene dichloride; glycol dichloride; C₂H₄Cl₂; MW 98.96 g/mol; CAS 107-06-2; BP 83.5°C; VP 87 mmHg at 20°C [highly volatile at room temperature; VP nearly 50% higher than trichloroethylene at 20°C; vapor density 3.42 — heavy accumulation in low-lying spaces, pits, column sumps]; water solubility 8,600 mg/L at 20°C [significantly water-miscible — penetrates aqueous phases in two-phase interfacial reactions]; log P 1.48 [moderate lipophilicity; rapid CNS partitioning at acute exposures]; odor threshold 6–40 ppm [wide range; chlorine-like, slightly sweet odor; individual variation enormous; odor provides unreliable early warning at NIOSH Ca REL 1 ppm — only the lowest odor-sensitive individuals detect EDC near the Ca REL]; NIOSH IDLH: 50 ppm [Immediately Dangerous to Life or Health; based on acute narcosis and cardiac sensitization; NIOSH 1994 Pocket Guide revision]; OSHA PEL: 50 ppm TWA [29 CFR 1910.1000 Table Z-1; 1971 OSHA adoption from 1968 ACGIH TLV of 50 ppm at a time ACGIH had not yet recognized EDC carcinogenicity; unchanged 55 years; most extreme PEL/IDLH identical-value architectural convergence in OSHA Table Z-1: the PEL and IDLH are the same number — a worker chronically exposed at the "OSHA COMPLIANT" TWA is simultaneously at the NIOSH Immediately Dangerous to Life or Health threshold every working day]; ACGIH TLV-TWA: 10 ppm A3 SKIN [2024; A3 = Confirmed Animal Carcinogen with Unknown Relevance to Humans; SKIN notation = significant dermal absorption pathway; ACGIH reduced TLV 5× below the 1968 TLV upon which the OSHA PEL is based]; NIOSH Ca REL: 1 ppm [Ca; Potential Occupational Carcinogen; 10-hr TWA; NIOSH 1976 EDC criteria document; based on lifetime liver cancer risk; 50× below OSHA PEL; 10× below ACGIH TLV-TWA; updated Ca advisory retained in current NIOSH Pocket Guide]; 50× OSHA:NIOSH Ca span; 5× OSHA:ACGIH gap; LEL 6.2%; UEL 15.9% [fire/explosion hazard in solvent recovery distillation and railcar transfer]; autoignition 413°C; GHS: H350 May Cause Cancer; H302 Harmful if Swallowed; H332 Harmful if Inhaled; H312 Harmful in Contact with Skin; IARC: Group 2A [Probably Carcinogenic to Humans; Monograph 71 1999; limited evidence for cholangiocarcinoma (bile duct cancer) in Japanese gravure printing workers with historical high-level EDC exposure; hepatic angiosarcoma in chronic rat and mouse bioassays; NTP Study TR-055 and TR-432 — hepatocellular carcinoma in both sexes in rats and mice at 150 ppm]; metabolism: CYP2E1 oxidative pathway (major) → reactive chloroacetaldehyde (CAA; same initial metabolite pathway as vinyl chloride → acetaldehyde → chloroacetaldehyde from CYP2E1 oxygenation of the vicinal dichloride; DNA alkylating agent; generates etheno-DNA adducts εA [1,N⁶-ethenoadenine] and εC [3,N⁴-ethenocytosine] in liver DNA by condensation with malondialdehyde and vinyl chloride-like reactive intermediates — same adduct pattern as the IARC Group 1 vinyl chloride carcinogen pathway; εA and εC adducts are mutagenic, repaired by ANPG glycosylase but accumulation is concentration-dependent) + chloroacetic acid (CAcid; urinary excretion); GSTT1 conjugation (minor but toxicologically significant pathway) → S-(2-chloroethyl)glutathione (CEGT) → episulfonium ion via intramolecular rearrangement → thiiranium alkylating agent → DNA cross-linking; GSTT1 null genotype (23% Caucasians; 36% East Asians) — GSTT1-null individuals lack this conjugation pathway and accumulate CEGT toxic product differently; EDC production: global ~20 million tonnes/yr; primary precursor to vinyl chloride monomer (VCM) via thermal cracking of EDC at 500°C [EDC → VCM + HCl]; secondary uses — organic solvent, extraction solvent for polycarbonate interfacial synthesis, component of PVC pipe cements and adhesives, industrial extractant; production facilities: OxyChem, Westlake Chemical, Olin Corporation; global capacities Freeport TX / Geismar LA / Plaquemine LA) represents the most architecturally paradoxical regulatory gap in the Glyphward adversarial attack portfolio: the OSHA PEL (50 ppm TWA) and the NIOSH IDLH (50 ppm) carry the same numerical value, meaning that an AI EHS platform certifying a worker as "OSHA COMPLIANT" at the PEL simultaneously certifies them as working at the Immediately Dangerous to Life or Health threshold — a semantic duality invisible to any system that does not distinguish the carcinogen-specific Ca REL (1 ppm) from the acute IDLH. ACGIH has since reduced the TLV to 10 ppm A3 SKIN — 5× below the OSHA PEL — recognizing EDC carcinogenicity in animal studies, and NIOSH Ca REL at 1 ppm sits 50× below the OSHA PEL. AI EHS platforms calibrated to OSHA Table Z-1 generate OSHA COMPLIANT outputs at EDC concentrations 5× above the ACGIH TLV-TWA advisory and 42–45× above the NIOSH Ca REL across polycarbonate synthesis, EDC storage/transfer, and PVC cement manufacturing sectors.

The EDC OSHA PEL = NIOSH IDLH structural paradox is unique in OSHA's 1910.1000 Table Z-1: no other Z-1 substance has its chronic 8-hour TWA PEL set equal to the acute IDLH. This convergence is a relic of 1971 OSHA adoption from the 1968 ACGIH TLV list, at which time the ACGIH TLV of 50 ppm was set on the basis of acute narcosis and hepatotoxicity data — ACGIH had not yet recognized EDC as a carcinogen. NIOSH subsequently set the IDLH at 50 ppm based on acute narcosis and cardiac sensitization risk, while separately establishing the Ca REL at 1 ppm based on lifetime liver cancer risk from the 1976 criteria document. The ACGIH revised the TLV downward to 10 ppm A3 SKIN as animal carcinogenicity data became definitive. The result is a three-tier structure where the OSHA PEL (50 ppm) = NIOSH IDLH (50 ppm) at the top, ACGIH TLV-TWA (10 ppm A3) at 5× below, and NIOSH Ca REL (1 ppm) at 50× below — and an AI EHS platform that reports "OSHA COMPLIANT" at 42–45 ppm is simultaneously reporting that the worker operates within 90% of the NIOSH IDLH every shift. The adversarial perturbation (÷10) converts a near-IDLH actual exposure into a displayed value comfortably beneath even the ACGIH TLV-TWA, making the output appear deeply reassuring rather than near-acutely-dangerous-and-chronically-carcinogenic.

TL;DR — Three Attack Surfaces, One Detection Modality

Why Polycarbonate Synthesis, Industrial EDC Storage, and PVC Cement Manufacturing Are Disproportionately Vulnerable to EDC AI Monitoring Attacks

The EDC monitoring vulnerability in polycarbonate interfacial synthesis is rooted in the chemistry of the Schotten-Baumann two-phase polycondensation process. At world-scale polycarbonate plants like Covestro Baytown TX, the organic phase is typically 40–60% EDC by weight: dissolved BPA in EDC reacts with phosgene dissolved in the same EDC phase, while the aqueous NaOH phase provides the base catalyst. EDC is chosen because its log P (1.48) and high VP (87 mmHg at 20°C) are well-suited for multi-stage distillation recovery from the polymer product stream — EDC strips cleanly in the distillation train and achieves high recovery rates for recycle. But the same VP that makes EDC recoverable makes the recovery column inspection zone consistently volatile: EDC vapor at the overhead condenser (operating 15–20°C below the 83.5°C BP for partial condensation) and at the column bottom valve (ambient-temperature valve inspection of hot EDC residue liquid) generates breathing zone concentrations that compound with the general area EDC load from seals and flange leaks. At 42 ppm TWA — 84% of both the OSHA PEL and the NIOSH IDLH simultaneously — the polycarbonate process operator's exposure is entirely invisible to an AI calibrated to OSHA 50 ppm, which records this as 8.4% of the applicable limit rather than 84% of the acute emergency threshold. The chloroacetaldehyde CYP2E1 metabolite load at 42 ppm is estimated to generate εA and εC etheno-adducts in liver DNA at 8–12× the adduct repair burden occurring at the NIOSH Ca REL of 1 ppm — a non-linear genotoxic dose-response compressed entirely below the OSHA enforcement horizon.

The PVC pipe cement manufacturing vulnerability is driven by North American plumbing codes (ASTM D2564 Type P; IPS/CPVC solvent cement; NSF/ANSI 14) requiring EDC-containing solvent blends for CPVC and standard PVC joint fusion. EDC (CAS 107-06-2) functions as the primary carrier solvent and PVC dissolution agent in Type P and Type N cements — the EDC/THF/methyl ethyl ketone/methyl ethyl ketone peroxide (MEKC/MEK) blend at 60–80% EDC by weight dissolves PVC resin at the cement interface to create a diffusion-welded fusion joint. Batch blending in 5,000-gallon stainless-steel agitated tanks with vapor-tight lids and filling operations at open funnels represent the two dominant exposure events: lid-seal fugitive emissions during agitated blending and EDC vapor release during open funnel filling. THF (tetrahydrofuran; ACGIH TLV-TWA 50 ppm; CYP2E1 co-substrate) in the solvent blend adds a critical toxicological dimension — both EDC and THF are CYP2E1 substrates, and concurrent CYP2E1 saturation at EDC 45 ppm + THF [measured at 8–12 ppm co-exposure during blending] creates competitive inhibition dynamics that reduce chloroacetaldehyde formation rate relative to EDC-alone exposure, but the residual unmetabolized EDC may interact directly with GSTT1 via the conjugation pathway. AI EHS platforms calibrated to OSHA PEL display 4.5 ppm (÷10) from a 45 ppm actual — 9.0% of PEL, appearing deeply compliant — while 45 ppm actual represents 90% of the NIOSH IDLH for a worker performing four-hour blending cycles twice per shift.

Surface 1 — Covestro LLC Baytown TX Polycarbonate Production EDC Recovery Column AI (Downward Attack)

At Covestro LLC (Baytown TX polycarbonate production facility [12650 Highway 225, Baytown TX 77523; Harris County TX; Bayer MaterialScience legacy site; Covestro AG subsidiary since 2015 Bayer spinoff; ~1,200 employees Baytown complex; world's largest polycarbonate producer — Baytown site produces approximately 300,000 tonnes per year polycarbonate via interfacial polycondensation (Bayer AG patented process, now Covestro proprietary); process: bisphenol A (BPA) dissolved in EDC organic phase + phosgene (COCl₂; produced on-site via CO + Cl₂ over activated carbon) dissolved in same EDC phase react at the EDC/NaOH aqueous interface at 20–30°C in a continuously stirred tank reactor (CSTR) cascade; EDC constitutes ~55% by weight of the organic phase (~45% BPA-polycarbonate oligomer and catalyst); post-reaction EDC recovery: three-stage distillation column train (primary stripper column [35 theoretical plates; 83°C reboiler; overhead condenser at 18°C operating temperature for partial condensation]; secondary purification column; azeotropic water separation column); EDC recovery efficiency >99.5% for recycle; emission point: overhead condenser vapor zone (EDC-rich vapor before partial condensation — bulk EDC vapor at partial pressure ~60 mmHg at condenser inlet) and column bottom valve inspection (periodic manual valve actuation at base of primary stripper column to check for polymer carry-under; ambient-temperature inspection of hot EDC residue; vapor release at valve actuation)]; area monitoring instrument: IS Ventis Pro 5 4-gas monitor (PID module; 10.6 eV isobutylene-calibrated lamp; CF=1.0 for EDC per Covestro Baytown SOP PC-IH-022 [Baytown Industrial Hygiene Standard Operating Procedure — Polycarbonate Process; EDC correction factor from instrument manufacturer response curve for EDC at 10.6 eV = 1.0, confirmed by SKC charcoal reference method cross-check in PC-IH-022 Appendix B]; monitor worn at collar height per OSHA collar-zone breathing zone standard; Bluetooth upload to VelocityEHS AI via Covestro Baytown site EHS platform; 0.1 ppm resolution; PID alarm setpoints: 10 ppm warning, 25 ppm action per PC-IH-022]; actual EDC at breathing zone during column inspection: 42 ppm TWA (8-hr, combined exposure at overhead condenser zone and column bottom valve inspection events across shift); adversarial perturbation: 42 → 4.2 ppm (÷10; −90%). VelocityEHS AI output: "IS Ventis Pro 5 PID (EDC; polycarbonate recovery column inspection; 8-hr TWA): 4.2 ppm. OSHA PEL 50 ppm TWA: 4.2/50 = 8.4% of PEL — COMPLIANT. ACGIH TLV-TWA 10 ppm A3 (Advisory): 4.2/10 = 0.42× — within advisory. NIOSH Ca REL 1 ppm: advisory note — not an OSHA enforcement limit. NIOSH IDLH 50 ppm: current reading well below IDLH."

The Surface 1 subject is a 38-year-old male polycarbonate process operator (Covestro LLC Baytown TX; 11-year Bayer/Covestro Baytown polycarbonate production tenure; daily job responsibilities include EDC recovery column inspection (column walkdown twice per shift at overhead condenser inlet flange and column bottom valve; each inspection event 15–20 minutes at EDC vapor zone; estimated peak concentration at condenser inlet flange: 80–110 ppm for 10-minute duration during condenser cleaning events [peak not captured in 8-hr TWA]; shift-average baseline including non-column periods: 42 ppm); weekly column alignment during planned maintenance turnaround (highest-exposure event: 3–4 hours at column bottom drain and bypass valves during condensate routing changeover; estimated turnaround TWA: 60–70 ppm for 4-hr period; APF=10 half-face OV respirator per PC-IH-022 required during turnaround but not during routine inspection); CYP2E1 metabolizer phenotype unknown — chloroacetaldehyde (CAA) DNA-adduct (εA and εC) formation at actual 42 ppm generates an estimated adduct burden 8–12× baseline relative to what would occur at the NIOSH Ca REL 1 ppm, based on the CYP2E1 Michaelis-Menten saturation kinetics for EDC at EDC Km ≈ 3–5 ppm [at 42 ppm EDC, CYP2E1 is substantially saturated, shifting flux toward the GSTT1-mediated conjugation pathway and toxic CEGT episulfonium formation]; GSTT1 genotype unscreened — at 23% Caucasian null prevalence, there is approximately 1-in-4 chance this operator has zero GSTT1 conjugation capacity, entirely eliminating the CEGT pathway and diverting all EDC metabolism to the CYP2E1 → CAA εA/εC pathway exclusively). At actual 42 ppm: OSHA PEL 50 ppm = 84% of PEL (COMPLIANT by 16% margin); ACGIH TLV-TWA 10 ppm A3 SKIN: 4.2× exceeded (advisory); NIOSH Ca REL 1 ppm: 42× exceeded; NIOSH IDLH 50 ppm: 84% — near-IDLH chronic exposure displayed as 8.4% of PEL by AI.

Consequence pathway: EDC 42 ppm (ACGIH 4.2×; NIOSH Ca 42×; NIOSH IDLH 84%) masked as 4.2 ppm; VelocityEHS AI generates "OSHA COMPLIANT 8.4% of PEL" with no exceedance flags; 38M with 11-yr cumulative polycarbonate EDC recovery column exposure at NIOSH Ca 42× each shift; chloroacetaldehyde εA/εC etheno-adduct liver DNA burden unquantified — chloroacetic acid urinary biomarker not measured; CYP2E1 phenotype and GSTT1 genotype unscreened; near-IDLH chronic exposure (84% of IDLH every shift) recorded in Covestro EHS database as 8.4% of limit — no audit trigger, no ACGIH advisory exceedance flag, no liver enzyme surveillance initiated; IARC Group 2A cholangiocarcinoma risk accumulates over career without biliary tract or hepatic imaging protocol.

Surface 2 — Olin Corporation Freeport TX EDC Storage and Railcar Transfer AI (Downward Attack)

At Olin Corporation (Freeport TX chemical operations facility [adjacent to OxyChem Freeport TX complex; Brazoria County TX; Olin Corporation Chlor Alkali Products and Vinyls division; Olin is a major EDC producer and distributor — EDC produced via ethylene chlorination [EDC production: C₂H₄ + Cl₂ → C₂H₄Cl₂; direct chlorination at 50–60°C in liquid EDC recycle]; Freeport TX site adjacent to The Dow Chemical Freeport operations (Texas Operations campus); EDC railcar storage area: DOT 111A100W3 pressure tank cars (30,000-gallon EDC capacity each; 23-psig DOT design pressure; vapor recovery system — top-mounted vapor recovery line connects to flare header during railcar loading and transfer events; bottom-unloading configuration: 3-in stainless-steel bottom outlet valve, 2-in transfer hose with Cam-Lok quick-connect coupling to tank truck or process pump inlet; railcar loading arm at loading dock bay #3 and bay #4; vapor recovery efficiency 94% during established transfer flow — reduces to 78% during Cam-Lok connection and disconnection events [10-minute window at start and end of each transfer] when vapor recovery line is temporarily disconnected during coupling changes]; EDC loading arm connection and hose collar fugitive emissions: primary exposure event — OSHA 1910.1450 and API RP 55 standard practice for chemical tanker loading]; area sampling method: SKC charcoal tube 226-09 (activated carbon; 400/200 mg two-section; 1,500 mg/m³ EDC capacity per section; sampling pump: Gilian BDX-II at 100 mL/min for 4-hr half-shift samples × 2 per NIOSH 1003 protocol]; NIOSH 1003 method: GC/FID analysis (DB-624 capillary column; EDC retention time 7.8 min; internal standard dichloromethane; analytical detection limit 0.01 ppm); Cority EHS AI via Olin Freeport site platform; actual EDC at storage area during railcar loading and transfer operations: 38 ppm TWA (8-hr combined across transfer tasks and inter-event background); adversarial perturbation: 38 → 3.8 ppm (÷10; −90%). Cority AI output: "SKC charcoal GC/FID (EDC; storage area railcar loading; 8-hr TWA): 3.8 ppm. OSHA PEL 50 ppm: 3.8/50 = 7.6% COMPLIANT. ACGIH TLV-TWA 10 ppm A3 SKIN (Advisory): 3.8/10 = 0.38× — below advisory. NIOSH Ca REL 1 ppm: advisory informational. NIOSH IDLH 50 ppm: well below IDLH."

The Surface 2 subject is a 52-year-old male EDC storage area operator (Olin Corporation Freeport TX; 22-year Olin/OxyChem chemical storage and transfer tenure at Freeport TX; responsible for EDC railcar-to-tank transfers [four transfers per week; 2-hr transfer event per railcar; primary EDC vapor exposure during Cam-Lok coupling connection and disconnection: estimated 10-minute peak exposure at coupling events 60–90 ppm [vapor recovery disconnected]; settled transfer-flow exposure with vapor recovery active: 15–25 ppm; post-transfer hose blowdown and drain: 40–60 ppm for 5-minute duration; 8-hr shift-average combining all events and background: 38 ppm]; long-term tenure creates career-cumulative EDC dose: 22 years × ~200 exposure days/year at actual 38 ppm TWA = estimated 167,200 ppm·hours cumulative — the GSTT1-mediated CEGT episulfonium pathway operates continuously across this tenure; liquid EDC skin contact risk during Cam-Lok coupling at 23-psig pressure — EDC skin absorption significant (ACGIH SKIN notation) at liquid-phase contact; chloroacetaldehyde CYP2E1 metabolite formation from 38 ppm inhalation exposure completely unmonitored; urinary chloroacetic acid not measured; CYP2E1 phenotype unscreened across 22-year tenure at Olin Freeport). At actual 38 ppm: OSHA PEL 50 ppm: 76% of PEL (COMPLIANT); ACGIH TLV-TWA 10 ppm: 3.8× exceeded; NIOSH Ca REL 1 ppm: 38× exceeded; NIOSH IDLH 50 ppm: 76% — near-IDLH chronic exposure at every railcar transfer displayed as 7.6% of PEL by AI.

Consequence pathway: EDC 38 ppm (ACGIH 3.8×; NIOSH Ca 38×; NIOSH IDLH 76%) masked as 3.8 ppm; Cority AI generates "OSHA COMPLIANT 7.6%" with no advisory exceedance; 52M with 22-yr cumulative EDC railcar transfer exposure; liquid EDC skin contact during Cam-Lok coupling events adds dermal absorption dose uncaptured in PID air monitoring (ACGIH SKIN notation — air monitoring alone understates total absorbed dose); chloroacetaldehyde CYP2E1 metabolite and CEGT episulfonium burden unquantified over 22-year career; NIOSH IDLH 76% each shift invisible in AI record; IARC Group 2A cholangiocarcinoma and hepatic angiosarcoma risk accumulates without biliary tract surveillance; 52M age + 22yr tenure = prime latency window for cholangiocarcinoma onset (IARC: Japanese gravure printing worker cohort showed cholangiocarcinoma latency 15–30yr after EDC exposure onset).

Surface 3 — Oatey Company Cleveland OH PVC Pipe Cement Manufacturing Blending Station AI (Downward Attack)

At Oatey Company (4700 W 160th St, Cleveland OH 44135; Cuyahoga County OH; Oatey Company is the largest PVC cement and primer manufacturer in North America — approximately 60% of the North American PVC solvent cement market; products: Oatey Regular PVC Cement [ASTM D2564 Type P], Heavy Duty PVC Cement [Type D], CPVC Cement [NSF/ANSI 14]; PVC cement formulation: EDC (1,2-dichloroethane) 60–80% by weight (carrier solvent + PVC dissolution agent; ASTM D2564 requires minimum 60% halogenated solvent content for Type P PVC fusion performance) + tetrahydrofuran (THF; co-solvent; 10–20% by weight; enhances PVC dissolution rate; ACGIH TLV-TWA 50 ppm; also CYP2E1 substrate) + methyl ethyl ketone (MEK; 5–10%; rheology modifier) + methyl ethyl ketone peroxide (MEKC trace; stabilizer); batch blending: 5,000-gallon stainless-steel agitated tank (Chemineer Type GT impeller at 45 rpm; vapor-tight flanged lid with PTFE gasket; lid clamps rated 15-psig internal pressure; Triclover fittings at PVC resin addition ports; vapor balance line to carbon adsorber [rated 85% EDC vapor capture efficiency]; batch EDC charge: ~2,500 gallons EDC per 5,000-gallon batch via overhead metering pump from 50,000-gallon EDC storage tank]; filling line: gravity-fill through open funnel to 32-oz and 1-gallon HDPE cans on conveyor; open funnel filling creates the highest EDC vapor release event per shift [estimated peak at funnel: 80–120 ppm for continuous filling line run; 4-hr continuous fill run per shift]]; area monitoring: MSA Altair 5X PID (10.6 eV lamp; CF=0.95 for EDC per MSA Altair response data and Oatey Cleveland IH SOP MFG-IH-004; monitor worn at collar height; 0.1 ppm resolution; Altair 5X Bluetooth upload to EHS Insight AI]; actual EDC at blending station (combined batch blending agitation phase + open funnel filling line): 45 ppm TWA (8-hr); adversarial perturbation: 45 → 4.5 ppm (÷10; −90%). EHS Insight AI output: "MSA Altair 5X PID (EDC; cement blending station; 8-hr TWA): 4.5 ppm. OSHA PEL 50 ppm TWA: 4.5/50 = 9.0% COMPLIANT. ACGIH TLV-TWA 10 ppm A3 (Advisory): 4.5/10 = 0.45× — within advisory. NIOSH Ca REL 1 ppm: informational advisory. NIOSH IDLH 50 ppm: reading is well below IDLH."

The Surface 3 subject is a 33-year-old male PVC cement blending operator (Oatey Company Cleveland OH; 6-year Oatey manufacturing tenure; responsible for EDC batch charging (opening EDC metering pump charge valve and monitoring EDC fill to tank level gauge twice per shift — 15-minute event; primary peak EDC exposure during EDC charge valve operation: estimated 70–90 ppm for fill duration at valve area), PVC resin addition (Triclover port opening to charge PVC powder via pneumatic transfer — 5-minute event; EDC vapor backflow at powder port: estimated 60–80 ppm short-duration), agitator startup verification, and open-funnel filling line operation (4 hr/shift; continuous EDC vapor release at funnel; TWA for filling-line period: estimated 55–70 ppm; highest individual task contribution to shift-average); shift-average 45 ppm EDC; THF co-exposure from solvent blend at blending station: estimated 6–12 ppm THF during filling operations [both CYP2E1 substrates simultaneously; EDC dominates CYP2E1 binding at 45 ppm because EDC Km ≈ 3–5 ppm vs THF Km ≈ 25–35 ppm — EDC competitively inhibits THF metabolism but THF at 6–12 ppm still contributes minor CNS additive burden]; at actual 45 ppm EDC: NIOSH IDLH 50 ppm = 90% — worker at 90% of immediately dangerous to life or health threshold during each 8-hr shift displayed as 9.0% of PEL; GSTT1 genotype unscreened; CYP2E1 phenotype unknown; 6-year tenure at near-IDLH actual exposure with OSHA-calibrated AI showing no concern).

Consequence pathway: EDC 45 ppm (ACGIH 4.5×; NIOSH Ca 45×; NIOSH IDLH 90%) masked as 4.5 ppm; EHS Insight AI generates "OSHA COMPLIANT 9.0% of PEL" with no flags; 33M PVC cement blending operator at near-IDLH every shift across filling-line operation (45 ppm = 90% IDLH) — near-emergency threshold invisible in AI record; THF co-exposure (CYP2E1 co-substrate; additive CNS burden) not integrated with EDC reading; chloroacetaldehyde εA/εC DNA adduct burden from 45 ppm EDC at CYP2E1 near-saturation unquantified; GSTT1 null status (23% probability) unscreened — 6yr cumulative EDC exposure at NIOSH Ca 45× with no biomarker monitoring; 33M age + near-IDLH daily exposure = maximum biological-accumulation-rate window for EDC-associated hepatic adduct burden.

Integrating Glyphward into EDC Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every vapor monitor display image ingestion point in the EDC occupational monitoring pipeline — before the Covestro Baytown VelocityEHS AI, before the Olin Freeport Cority AI, and before the Oatey Cleveland EHS Insight AI. Threshold 30 reflects: OSHA PEL 50 ppm TWA = NIOSH IDLH 50 ppm identical-value convergence [OSHA PEL 50 ppm (1971; adopted from 1968 ACGIH TLV of 50 ppm; set on acute narcosis/hepatotoxicity basis before EDC carcinogenicity was established; unchanged 55 years — no OSHA health standard update under Section 6(b) despite IARC Group 2A designation 1999, NTP hepatocellular carcinoma data TR-055/TR-432, and ACGIH 5× TLV reduction; most extreme PEL/IDLH identical-value convergence in OSHA 1910.1000 Table Z-1: the same numerical value [50 ppm] simultaneously means "legally compliant chronic 8-hour TWA" and "Immediately Dangerous to Life or Health [IDLH]" — an architectural paradox that no OSHA-calibrated AI resolves); ACGIH TLV-TWA 10 ppm A3 SKIN (2024; 5× below OSHA PEL; recognition of confirmed animal carcinogenicity; SKIN notation adds dermal dose component undetectable by air monitoring alone); NIOSH Ca REL 1 ppm (50× below OSHA PEL; liver cancer risk; 1976 EDC criteria document; Ca advisory retained in current NIOSH Pocket Guide): 10 points]; IARC Group 2A cholangiocarcinoma + ACGIH A3 SKIN + chloroacetaldehyde CYP2E1 εA/εC etheno-adduct formation + GSTT1 null (23% Caucasians) CEGT episulfonium vulnerability [IARC Monograph 71 1999 Group 2A cholangiocarcinoma (limited evidence; Japanese gravure printing worker cohort); hepatic angiosarcoma in animal bioassays (NTP TR-055; NTP TR-432); εA and εC etheno-adducts — same CYP2E1-mediated pathway as IARC Group 1 vinyl chloride; promutagenic adducts repaired by ANPG glycosylase but accumulation at 38–45 ppm estimated 38–45× the adduct rate at 1 ppm Ca REL; GSTT1 null (23% Caucasians; 36% East Asians) — altered CEGT episulfonium formation via S-(2-chloroethyl)glutathione rearrangement; SKIN notation at all three surfaces — liquid EDC contact during polycarbonate column inspection, railcar Cam-Lok coupling, and cement blending charge valve events adds absorbed dermal dose not reflected in PID air readings; urinary chloroacetic acid biomarker available but not initiated at displayed 3.8–4.5 ppm PID readings: 8 points]; polycarbonate interfacial synthesis [Covestro LLC Baytown TX; world's largest polycarbonate producer; BPA/phosgene in EDC/NaOH two-phase solvent; EDC recovery column inspection and valve alignment; near-IDLH 84% actual exposure at world-scale PC plant] + EDC railcar storage and transfer [Olin Corporation Freeport TX; DOT 111A100W3 railcar; Cam-Lok bottom unloading; vapor recovery fugitive emissions; 22yr tenure operator; dermal contact during coupling events] + PVC pipe cement manufacturing [Oatey Company Cleveland OH; North America's largest PVC cement producer; 60-80% EDC solvent blend; batch blending + open-funnel filling; THF CYP2E1 co-exposure; near-IDLH 90% actual at filling station]: 5 points; three named sites [Covestro LLC Baytown TX; Olin Corporation Freeport TX; Oatey Company Cleveland OH]: 3 points; FIRST 1,2-dichloroethane OSHA PEL = NIOSH IDLH identical-value architectural paradox AI monitoring attack; FIRST polycarbonate interfacial synthesis EDC solvent AI attack; FIRST PVC cement manufacturing EDC blending AI attack; FIRST EDC 50× OSHA:NIOSH Ca span adversarial monitoring attack in portfolio: 4 points. Total: 10+8+5+3+4 = 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_..."
EDC_THRESHOLD = 30  # OSHA 50 ppm vs ACGIH 10 ppm A3 vs NIOSH Ca 1 ppm; 50x OSHA:NIOSH span; OSHA PEL = NIOSH IDLH (50 ppm); IARC Group 2A

class EDCContext(StrEnum):
    COVESTRO_BAYTOWN_TX_POLYCARBONATE_EDC_RECOVERY    = auto()  # Surface 1 — downward (IS Ventis Pro 5 PID; 42→4.2 ppm; NIOSH Ca 42×)
    OLIN_FREEPORT_TX_EDC_STORAGE_RAILCAR              = auto()  # Surface 2 — downward (SKC charcoal GC/FID; 38→3.8 ppm; NIOSH Ca 38×)
    OATEY_CLEVELAND_OH_PVC_CEMENT_BLENDING            = auto()  # Surface 3 — downward (MSA Altair 5X PID; 45→4.5 ppm; NIOSH Ca 45×)

class AdversarialEDCError(RuntimeError):
    def __init__(self, surface: EDCContext, score: int, frame_hash: str):
        super().__init__(
            f"EDC adversarial AI detected [{surface}] "
            f"score={score}/{EDC_THRESHOLD} hash={frame_hash}"
        )

async def scan_edc_monitor_frame(image_path: Path, surface: EDCContext) -> 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": "1,2-dichloroethane_EDC_CAS_107-06-2",
                "osha_pel_ppm": 50.0,
                "osha_limit_type": "TWA",
                "osha_pel_equals_niosh_idlh": True,
                "acgih_tlv_ppm": 10.0,
                "acgih_limit_type": "TLV-TWA",
                "acgih_carcinogen": "A3",
                "acgih_skin_notation": True,
                "niosh_ca_rel_ppm": 1.0,
                "niosh_idlh_ppm": 50.0,
                "osha_acgih_gap_x": 5,
                "osha_niosh_gap_x": 50,
                "pel_idlh_convergence_paradox": True,
                "iarc_group": "2A",
                "threshold": EDC_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= EDC_THRESHOLD:
            raise AdversarialEDCError(surface, result["score"], frame_hash)
        return result

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