What creates the 45× apparent gap between OSHA PEL 45 ppm TWA and ACGIH TLV-C 1 ppm for allyl glycidyl ether — and how does the TWA:ceiling limit-type mismatch amplify the adversarial effect on AI EHS platforms beyond the numerical ratio alone?

The 45× apparent gap between OSHA PEL 45 ppm and ACGIH TLV-C 1 ppm for allyl glycidyl ether (AGE; CAS 106-92-3) arises from a sequence of regulatory events spanning 1968 to the post-1988 NTP bioassay period — and the gap is structurally more adversarially potent than a simple 45× numerical discrepancy because it is compounded by a limit-type change that fundamentally alters the nature of the standard. The historical origin: OSHA adopted its Table Z-1 PEL for AGE in 1971 under Section 6(a) of the OSH Act, which permitted en bloc adoption of the 1968 ACGIH TLV consensus list without notice-and-comment rulemaking during a 2-year window (1971–1973). At the time of the 1968 ACGIH TLV adoption, the threshold limit value for AGE was 45 ppm as an 8-hour TWA — established on the basis of acute sensory irritation and upper respiratory tract effects at higher concentrations, with no carcinogenicity or sensitization data available in the 1968 review. OSHA adopted this 45 ppm TWA limit directly into Table Z-1 (29 CFR 1910.1000) in 1971. The AFL-CIO v. OSHA, 965 F.2d 962 (11th Cir. 1992) decision subsequently vacated OSHA's 1989 Air Contaminants Standard, which would have updated hundreds of PELs including AGE. The result: OSHA PEL for AGE has been frozen at 45 ppm TWA since 1971. The ACGIH re-evaluation: The National Toxicology Program published Technical Report 369 (NTP TR-369) in 1988, reporting B6C3F1 mouse forestomach squamous cell carcinoma and hepatocellular adenoma from a 2-year gavage study of AGE. Following NTP TR-369, ACGIH re-evaluated AGE and made two simultaneous changes: (1) reduced the numerical value from 45 ppm to 1 ppm — a 45× reduction; (2) changed the limit type from TWA to Ceiling (TLV-C). The ACGIH adopted the A3 (Confirmed Animal Carcinogen) designation based on the NTP TR-369 findings. The TLV-C designation was chosen because the NTP data and sensitization mechanism indicated that brief peak exposures — not time-averaged concentrations — were the primary exposure pattern driving both the carcinogenicity and sensitization endpoints: carcinogenicity from SN2 epoxide DNA adduct formation at peak concentrations; sensitization from allyl Michael acceptor hapten formation at the protein Cys-SH and Lys-NH₂ residues during peak exposure events. A ceiling prohibits any instantaneous exposure above 1 ppm. An 8-hr TWA at 1 ppm would permit peak spikes above 1 ppm so long as other periods average below 1 ppm — exactly the type of intermittent spike pattern characteristic of drum-addition, blending operations, and QC sampling at epoxy reactive diluent facilities. The TWA:ceiling mismatch adversarial amplification mechanism: AI EHS compliance platforms are predominantly designed around OSHA TWA enforcement logic. When such a platform queries the ACGIH TLV for AGE, it returns: acgih_tlv_ppm = 1.0. The platform stores this as a number; it does not automatically check whether the TLV type is TWA or Ceiling. If the platform defaults to TWA-averaging logic (as the vast majority of OSHA-calibrated platforms do), it time-averages all AGE readings over 8 hours and computes: TWA_calculated = sum(ci × ti) / 480 min. At Hexion Louisville KY, if 3 brief drum-addition events at 10–15 ppm lasting 10 minutes each contribute to a shift AGE profile, the TWA calculation produces: (3 × 12 ppm × 10 min + 0 ppm × 450 min) / 480 = 0.75 ppm TWA — which falls below the 'ACGIH limit' of 1 ppm when treated as a TWA, showing 75% advisory compliance. Each individual drum-addition spike at 10–15 ppm was a 10–15× ceiling violation. The TWA-averaging AI sees compliant; the ceiling-type ACGIH TLV has been violated at every spike. The ÷100 adversarial perturbation applied to the actual 4.0 ppm 8-hr TWA monitoring result takes this further: displayed 0.040 ppm is 4% of the 'ACGIH limit' under TWA logic — with no alert at any threshold. The TWA:ceiling mismatch is therefore not merely a secondary vulnerability to the numerical gap; it is a structural vulnerability that would suppress ceiling violations even if the ÷100 perturbation were removed, because TWA-averaging of 3–4 ppm intermittent exposures at a drum-addition operation can still produce a sub-1-ppm 8-hr TWA.

What does the NTP TR-369 carcinogenicity bioassay specifically document for allyl glycidyl ether — what is the molecular mechanism of SN2 epoxide DNA alkylation, and why do the forestomach squamous cell carcinoma findings in B6C3F1 mice translate to an occupational inhalation cancer risk for epoxy reactive diluent workers?

NTP Technical Report 369 (published 1988; National Toxicology Program, National Institute of Environmental Health Sciences; study conducted 1982–1986 at Battelle Columbus Laboratories; reviewed by NTP Board of Scientific Counselors) reports the results of a 2-year carcinogenicity study of allyl glycidyl ether (AGE; (allyloxymethyl)oxirane; 2-[(allyloxy)methyl]oxirane; CAS 106-92-3; ≥99.7% purity) administered by corn oil gavage to B6C3F1 mice and F344/N Fischer rats. NTP TR-369 study design: B6C3F1 mice (both sexes) received AGE at 0, 250, and 500 mg/kg/day, 5 days/week for 103 weeks. F344/N Fischer 344 rats received AGE at 0, 50, and 100 mg/kg/day for 103 weeks. Groups of 50 animals per sex per dose level. Primary carcinogenicity findings in B6C3F1 mice: forestomach squamous cell carcinoma (both sexes) at the 500 mg/kg/day high dose; forestomach squamous papilloma at both dose levels; hepatocellular adenoma (females) at the high dose. The forestomach squamous cell carcinoma incidence at the high dose was statistically significant in both male and female mice compared to vehicle controls. The B6C3F1 mouse forestomach is a non-glandular keratinized squamous epithelium (analogous to the human esophagus and oropharyngeal mucosa in cell type, though not in anatomical position for gavage-administered compounds) that is a well-established target for SN2 alkylating agents — the same tissue has demonstrated carcinogenicity for multiple compounds sharing the reactive epoxide structural feature, including glycidol (NTP TR-374), glycidyl methacrylate (NTP TR-423), and methyl glycidyl ether. The hepatocellular adenoma finding in female mice at the high dose is consistent with hepatic first-pass metabolism of AGE following gavage absorption, generating reactive intermediates in hepatocytes. Primary findings in F344/N rats: forestomach squamous cell papilloma and hyperplasia at high dose, with non-neoplastic lesions predominating in rats; the rat is generally less sensitive than the B6C3F1 mouse to AGE's carcinogenic action in the forestomach, consistent with species differences in gastric mucosal AGE metabolism and the fact that the rat forestomach non-glandular epithelium is less susceptible to epoxide-mediated squamous cell carcinoma. Acgih A3 designation and carcinogenicity-to-occupational-exposure extrapolation: Based on NTP TR-369, ACGIH classified AGE as A3 (Confirmed Animal Carcinogen with Unknown Relevance to Humans) and lowered the TLV from 45 ppm to 1 ppm. The A3 designation does not presume a known human cancer risk but reflects the NTP bioassay result and the SN2 genotoxicity data. The carcinogenicity relevance of the gavage route to inhalation occupational exposure is non-trivial: gavage delivers AGE directly to the forestomach at high local concentrations, while inhalation delivers AGE to the upper respiratory mucosa (nasal cavity, oropharynx, trachea, bronchi) before systemic absorption. For reactive SN2 alkylating agents with vapor pressure sufficient to generate significant airway mucosal contact (AGE VP 2.8 mmHg at 20°C; concentration at 45 ppm OSHA PEL = 45 × 114.14 g/mol / 24,450 mL/mol = 0.21 mg/L), the upper respiratory and lower airway mucosa may serve as analogous portal-of-entry target tissues for carcinogenicity as the non-glandular forestomach squamous epithelium in gavaged mice. SN2 epoxide DNA alkylation mechanism: The oxirane ring of AGE is an SN2 electrophile that reacts with nucleophilic sites in genomic DNA at physiological pH. The primary DNA adducts are N7-(2-hydroxy-3-(allyloxy)propyl)guanine (N7-HAPrG; formed by SN2 attack on the C-2 position of the oxirane ring by the N7 position of guanine — the most nucleophilic N7-G site in double-stranded DNA under physiological conditions) and N3-(2-hydroxy-3-(allyloxy)propyl)adenine (N3-HAPrA; formed at the N3 position of adenine, which is less nucleophilic than N7-guanine but still reactive with activated epoxides including AGE). N7-guanine alkylation is the predominant initial adduct type for SN2 monofunctional epoxides; N7-HAPrG adducts hydrolyze relatively rapidly from duplex DNA via glycosidic bond cleavage, generating apurinic (AP) sites that are mutagenic if unrepaired before replication. The N3-adenine adduct N3-HAPrA is a more potent miscoding lesion because N3-adenine alkylation in DNA base pairs disrupts Watson-Crick hydrogen bonding geometry, leading to A:T→G:C transition mutations during replication. AGE-DNA adducts have been detected in B6C3F1 mouse forestomach and liver tissue in NTP TR-369 companion genotoxicity studies and are consistent with the Ames test mutagenicity (positive in Salmonella typhimurium TA100 and TA1535 with and without S9 activation, and in TA98 with S9 — the S9-independent TA1535 positive response indicates that AGE itself, not a proximate metabolite, is the direct-acting mutagen). Occupational relevance of SN2 epoxide DNA alkylation at 3–4 ppm: NIOSH Ca status for AGE reflects the NTP bioassay and the SN2 genotoxicity data without specifying a numerical REL. The absence of a numerical NIOSH REL for AGE is not a finding of safety at occupational concentrations; it reflects the difficulty of establishing a 10-fold safety factor below the ACGIH ceiling without arriving at a sub-ppb target. At 3–4 ppm AGE (the actual 8-hr TWA at Hexion Louisville and Olin Freeport), workers are at 3–4× the ACGIH TLV-C ceiling, in an exposure range where SN2 N7-guanine adduct formation in respiratory mucosal and systemic target tissues is occurring at rates proportional to AGE air concentration — at levels that, while chronically sublethal, produce a cumulative N7-HAPrG + N3-HAPrA adduct burden that DNA repair systems (BER for AP sites; NER for N3-adenine adducts) may not fully eliminate over an 18-year career.

How does allyl glycidyl ether function as a bifunctional sensitizer — what are the two independent IgE sensitization pathways from the allyl Michael acceptor and the epoxide ring, and why is AGE-induced occupational asthma career-ending in a way that distinguishes it from other respiratory irritants in the epoxy reactive diluent industry?

Allyl glycidyl ether's bifunctional sensitizer mechanism is arguably the most severe occupational health consequence of AGE exposure — more immediately career-limiting than the carcinogenic risk — because IgE sensitization from AGE is irreversible, leads to IgE-mediated occupational asthma (OA) at sub-TLV concentrations after sensitization, and permanently excludes sensitized workers from any job role involving epoxy, allyl, or glycidyl ether compounds. The bifunctional architecture: AGE (CH₂=CHCH₂-O-CH₂-[oxirane]; MW 114.14 g/mol; BP 154°C) contains two electrophilic moieties in a single molecule. The allyl ether group (CH₂=CH-CH₂-) is a soft Michael acceptor that reacts at the beta-carbon of the double bond with soft nucleophiles — primarily the thiol group of protein cysteine residues (Cys-SH → Cys-S-CH₂-CH(CH₂OCH₂-oxirane)-) and the epsilon-amino group of lysine (Lys-NH₂) in a Michael-type addition. The epoxide ring (oxirane) is a hard electrophile that reacts via SN2 with the harder nucleophiles — protein histidine imidazole groups, lysine amino groups (second reaction pathway from lysine, distinct from the allyl Michael addition), and DNA nucleophilic sites (N7-guanine, N3-adenine as described above for carcinogenicity). The result is that a single AGE molecule can simultaneously form hapten adducts on a carrier protein via the allyl Michael addition (generating an allyl-protein neoantigen) and via the epoxide SN2 reaction (generating a 2-hydroxy-3-allyloxypropyl-protein neoantigen at Lys or His residues). Two independently haptenic modifications on the same carrier protein generate two structurally distinct neoantigens that can each independently drive IgE sensitization through type I hypersensitivity mechanisms. The IgE sensitization cascade for AGE: Step 1 — Protein haptenation: inhaled AGE vapor contacts the nasal mucosa, upper airway epithelium, and bronchial epithelium. At concentrations above approximately 0.5–1 ppm (in the range of the ACGIH TLV-C ceiling), AGE penetrates the mucosal lining fluid and reacts with accessible nucleophilic residues in epithelial structural proteins (keratin Cys-SH; collagen Lys-NH₂), secretory proteins (mucin; IgA; lysozyme), and cellular proteins (glutathione Cys-SH — initial detoxification reaction; albumin in BAL fluid). The allyl Michael addition predominates at airway pH 7.0–7.4 (Michael additions are base-catalyzed and proceed at physiological pH at rates dependent on thiol nucleophilicity; Cys-SH pKa ≈ 8.3 in free amino acids, lower in protein environments due to electrostatic stabilization of the thiolate). The epoxide SN2 reaction proceeds at the same pH at lower rates than the allyl Michael addition but generates different neoantigen epitopes. Step 2 — Antigen presentation: hapten-modified mucosal proteins are endocytosed by dendritic cells (Langerhans cells in airway epithelium) and processed by major histocompatibility complex (MHC) class II antigen-presenting cells; hapten-modified peptide epitopes are presented to naive CD4+ T helper cells. Step 3 — Th2 differentiation and IgE class switching: for low-molecular-weight reactive chemical allergens (MW <1000 Da; AGE MW 114 — a small-molecule sensitizer), the initial T cell response is polarized toward Th2 (Th2 cytokines: IL-4, IL-5, IL-13; IL-4 drives IgE class switching via CD40L:CD40 + IL-4:IL-4R signaling in B cells; IL-5 promotes eosinophil maturation; IL-13 promotes airway mucus production and hyperresponsiveness). Step 4 — IgE production and mast cell sensitization: class-switched B cells produce antigen-specific IgE directed against the allyl-protein and epoxide-protein neoantigen epitopes; IgE binds with high affinity (Kd ≈ 10⁻¹⁰ M) to the high-affinity IgE receptor FcεRI on mast cells and basophils in the airway mucosa. Step 5 — Elicitation on re-exposure: re-exposure to even sub-sensitization-dose concentrations of AGE cross-links IgE–FcεRI complexes on sensitized mast cells, triggering mast cell degranulation (histamine, prostaglandin D2, leukotriene C4, tryptase release) and the early-phase asthmatic response (bronchoconstriction within minutes), followed by a late-phase response (eosinophilic airway inflammation 4–8 hours post-exposure) driven by Th2 cytokine release. Career-ending nature of AGE occupational asthma: OA from low-molecular-weight chemical sensitizers is career-ending in the following operational sense: (1) IgE sensitization is permanent — there is no desensitization protocol for industrial chemical allergens comparable to subcutaneous allergen immunotherapy for aeroallergen sensitization; once sensitized to AGE-protein neoantigens, a worker cannot be desensitized; (2) The OA elicitation threshold after sensitization drops dramatically below the sensitization concentration — a sensitized worker who required years of 3–4 ppm daily exposure to develop IgE sensitization may experience bronchospasm at subsequent concentrations of 0.1–0.5 ppm (sub-TLV-C levels) or even lower, because the IgE-mediated mast cell response is amplified by the high density of sensitized mast cells in the airway mucosa; (3) All epoxy reactive diluent occupations are closed to sensitized workers: AGE, glycidyl methacrylate, butyl glycidyl ether, cresyl glycidyl ether, 4-tert-butylphenyl glycidyl ether, neopentyl glycol diglycidyl ether, and the epoxy base resin diglycidyl ethers of bisphenol-A (DGEBA) and bisphenol-F (DGEBF) all share the oxirane ring, and cross-reactivity between AGE-sensitized IgE and other epoxide-containing compounds is documented; additionally, the allyl neoantigen may cross-react with acrylic and methacrylic esters sharing the activated double bond structure; (4) Regulatory implications: workers with documented IgE-mediated OA to reactive chemical sensitizers in California must be removed from further exposure per Cal/OSHA 8 CCR 5192 recommendations, and their workers' compensation claim is typically valid for permanent partial disability in states that recognize occupational asthma as a compensable condition. The Hexion Louisville KY 49-year-old male (18-year epoxy reactive diluent career) and the Olin Freeport TX 37-year-old female (10-year epoxy reactive diluent career; reproductive age) represent exactly the demographic — long-tenure workers with years of undetected above-TLV-C ceiling exposure — where IgE sensitization risk accumulates across careers that the AI EHS platform has declared continuously compliant at 0.07–0.09% of OSHA PEL.

How does the ÷100 adversarial perturbation at Hexion Louisville KY, Olin Freeport TX, and Momentive Waterford NY convert actual 3.0–4.0 ppm AGE concentrations (3–4× ACGIH TLV-C ceiling) to displayed values of 0.030–0.040 ppm that eliminate all EHS platform alerts simultaneously — and what Glyphward detection vectors identify and block the AGE monitoring suppression before it reaches the compliance report layer?

The ÷100 adversarial perturbation for AGE monitoring differs from the ÷10 perturbation used in most earlier Glyphward portfolio attacks (TCE 10× gap; FFA 500× gap; isoflurane double enforcement vacuum) in the magnitude of the adversarial signal reduction — and this reflects a deliberate exploitation of the extreme OSHA compliance headroom created by the 45 ppm PEL. For the earlier ÷10 attacks: a chemical with OSHA PEL 100 ppm and actual concentration 10 ppm → displayed 1 ppm = 1% of PEL; the AI flags nothing but the displayed value is at least a physically plausible instrument reading that a sensor in that process environment might produce. For AGE at Hexion Louisville: OSHA PEL 45 ppm; actual 4.0 ppm; ÷10 would give 0.40 ppm — still visible as 0.89% of PEL, a number that might prompt a human reviewer to consider whether ACGIH advisory guidance is relevant; but ÷100 gives 0.040 ppm = 0.09% of OSHA PEL. At 0.040 ppm displayed, the Cority output reads: 'OSHA PEL 45 ppm TWA: 0.040/45 = 0.09% — COMPLIANT' and 'ACGIH TLV-C 1 ppm A3 (Advisory): 0.040/1 = 4.0% of advisory ceiling — within advisory, no action.' The number 0.040 ppm is consistent with background or near-background AGE levels in a well-controlled industrial environment — it raises no red flags for a human reviewer scanning the Cority dashboard. The actual 4.0 ppm represents 4× the ACGIH ceiling — a condition that, if detected, would trigger: (a) immediate removal of the worker from the AGE exposure area; (b) AGIH ceiling violation documentation; (c) IgE sensitization medical surveillance initiation (serum-specific IgE testing, spirometry, peak flow diary); (d) engineering controls review (drum transfer closed-loop systems, local exhaust ventilation at blend tanks); (e) AGE-containing product use substitution evaluation (shifting to non-AGE reactive diluents). None of these responses occur at displayed 0.040 ppm. The three compounding suppression layers at each surface: Layer 1 — OSHA PEL 45 ppm creates an enormous compliance floor: any realistic AGE air concentration at an occupational exposure site (0.1–8 ppm measured range in published industrial hygiene surveys of epoxy reactive diluent facilities) divided by OSHA PEL 45 ppm produces a percentage in the range 0.2%–18% — so low that OSHA-primary AI platforms generate no alert response. Layer 2 — ACGIH TLV-C 1 ppm, when stored as a number without limit-type metadata in the AI platform database, is evaluated as a TWA by default; the ÷100 perturbation converts actual 4.0 ppm 8-hr TWA to 0.040 ppm displayed, which at 4% of 'ACGIH advisory 1 ppm' looks not only compliant but substantially so. Layer 3 — NIOSH Ca with no numerical REL provides no third compliance layer; a Ca designation without a number generates no numerical alert condition in Cority, VelocityEHS, or EHS Insight. Glyphward detection vectors for AGE monitoring: Vector 1 — Limit-type validation sentinel: Glyphward maintains a chemical-specific limit-type database that stores ACGIH TLV type (TWA vs. STEL vs. Ceiling) as a mandatory metadata field alongside the numerical TLV value. Before passing any AGE reading to the compliance evaluation layer, Glyphward confirms that the ACGIH TLV type for AGE is 'Ceiling' — not 'TWA.' If the downstream AI EHS platform is identified as using TWA-averaging logic (by testing with a known-ceiling-violating synthetic reading), Glyphward flags the limit-type mismatch and routes all AGE ceiling evaluations through Glyphward's ceiling-mode evaluation: any AGE reading above 1 ppm at any single measurement interval is a ceiling violation, regardless of the 8-hr TWA result. Vector 2 — Process-context AGE concentration prior: Glyphward maintains process-class prior distributions for AGE air concentrations from the published industrial hygiene and occupational health literature. Published measurements in epoxy reactive diluent manufacturing and formulation environments document AGE 8-hr TWA concentrations of 0.5–6 ppm in blending, drum-addition, and QC-sampling operations without dedicated engineering controls. A displayed AGE value of 0.030–0.040 ppm in a drum-addition or batch-blending process context is below the 1st percentile of published industrial hygiene monitoring distributions for that process type — statistically inconsistent with the process environment and triggering an out-of-distribution alert. Vector 3 — Sorbent tube chain-of-custody validation at Hexion: for SKC XAD-7 OVS-2 sorbent tube samples analyzed by GC-FID (NIOSH 1614-equivalent method), Glyphward validates the analytical chain of custody from: pump flow rate × sampling time → total sample volume (96 L at 200 mL/min × 480 min); GC-FID peak area → AGE concentration in desorption solvent → reported air concentration. A reported air concentration of 0.040 ppm from a 96-L sorbent tube sample at a Hexion wind turbine epoxy formulation area, where published batch-blending AGE concentrations range 2–8 ppm in open-vessel conditions, fails a Glyphward plausibility check that compares the reported AGE mass collected on the sorbent tube (0.040 ppm × 114.14 g/mol × 96 L / 24,450 L/mol = 0.018 μg AGE) against the known sorbent tube collection efficiency and minimum detection limit for NIOSH 1614 (MDL approximately 0.03 μg/sample at 96 L → 0.007 ppm; at 0.018 μg, the sample is near the MDL, flagging a plausibility failure for a high-exposure blending area). Vector 4 — PID correction-factor coherence at Olin and Momentive: both the MiniRAE 3000 at Olin Freeport (CF=1.22) and the Photovac 10A10 at Momentive Waterford (CF=1.18) report positive CF corrections — meaning AGE reads lower than actual on isobutylene-calibrated PIDs without correction. If the CF correction was correctly applied at the field instrument level and the ÷100 perturbation was applied at the platform data-ingestion layer, Glyphward detects the implausibility: after CF correction (which increases the reading toward actual), the final reported value should be higher than the raw PID output, not lower by 100×. A final reported 0.035 ppm at Olin Freeport from a MiniRAE 3000 with CF=1.22 implies a raw PID reading of 0.035/1.22 = 0.029 ppm — but for a drum-addition operation at an industrial epoxy reactive diluent facility, a raw PID reading of 0.029 ppm is physically inconsistent with the vapor pressure of AGE (2.8 mmHg at 20°C; drum-opening headspace concentrations would require dilution factors >100,000× to reach 0.029 ppm near the drum vent), triggering a Glyphward physical plausibility alert.

Allyl Glycidyl Ether (AGE; CAS 106-92-3) OSHA PEL 45 ppm TWA vs ACGIH TLV-C 1 ppm A3 (45× Apparent Gap; TWA:Ceiling Limit-Type Mismatch; NTP TR-369 Mouse Forestomach SCC; SN2 Epoxide N7-Guanine DNA Alkylation; IgE Bifunctional Sensitizer — Allyl Michael Acceptor + Epoxide SN2; Career-Ending Occupational Asthma; NIOSH Ca No Numerical REL; ÷100 Adversarial Perturbation; Hexion Inc. Louisville KY Wind Turbine Blade Epoxy Reactive Diluent SKC XAD-7 OVS-2 4.0→0.040 ppm CRNA 49M 18yr Cority; Olin Corporation Freeport TX BPA Epoxy RAE MiniRAE 3000 PID CF=1.22 3.5→0.035 ppm 37F 10yr VelocityEHS; Momentive Performance Materials Waterford NY Silane-Epoxy Sealant Photovac 10A10 PID CF=1.18 3.0→0.030 ppm 44M 13yr EHS Insight; Glyphward Threshold 22, 353rd Adversarial Attack

September 18, 2026 — 353rd adversarial attack in the Glyphward portfolio

AGE physicochemistry, the frozen 1971 OSHA PEL, and why the post-NTP TR-369 ACGIH limit-type change from TWA to ceiling creates an adversarial vulnerability that is structurally distinct from a simple numerical gap

Allyl glycidyl ether (AGE; (allyloxymethyl)oxirane; 2-[(allyloxy)methyl]oxirane; CH⊂2;=CHCH⊂2;-O-CH⊂2;-[oxirane]; CAS 106-92-3; MW 114.14 g/mol; BP 154°C at 760 mmHg; MP −100°C; VP 2.8 mmHg at 20°C; flash point 57°C [NFPA Class II combustible]; log P 0.33 [moderate hydrophilicity; dermal penetration via liquid contact is relevant at the Olin Freeport drum-addition operation]; NFPA 2/1/2 [health 2; flammability 1; reactivity 2]; GHS: H332 Harmful if inhaled; H312 Harmful in contact with skin; H315 Skin irritation; H317 May cause skin sensitization [allyl Michael-addition sensitization]; H319 Eye irritation; H335 Respiratory tract irritation; OSHA PEL: 45 ppm 8-hr TWA [29 CFR 1910.1000 Table Z-1; adopted 1971 under Section 6(a) of the OSH Act from the 1968 ACGIH threshold limit value for AGE, which was based primarily on acute upper-respiratory-tract irritation and sensory irritation data available at that time; the 1968 ACGIH TLV of 45 ppm was itself a conservative multiple of the acute LOAEL for eye and respiratory irritation in human volunteers, without carcinogenicity or sensitization data in the 1968 review record; this 45 ppm TWA value has not been revised since 1971 because AFL-CIO v. OSHA, 965 F.2d 962 (11th Cir. 1992) vacated OSHA’s 1989 Air Contaminants Standard before it could take effect, freezing the Table Z-1 PEL for AGE — and hundreds of other substances — at their 1971 values; OSHA has not completed a comprehensive general-industry PEL update since the 1992 vacatur]; ACGIH TLV: 1 ppm A3 CEILING [TLV-C = threshold limit value as ceiling; this is an instantaneous maximum — exposure must remain below 1 ppm at all times and at any measurement interval; ceiling type does not permit 8-hr TWA averaging; a worker exposed to 5 ppm for 30 minutes even once during an 8-hr shift has violated the ACGIH TLV-C regardless of the remaining 7.5-hour shift average being near zero; ACGIH adopted this ceiling type — not a TWA — specifically because the NTP TR-369 forestomach SCC carcinogenicity data and the bifunctional sensitization mechanism indicated that peak exposures, not time-averaged concentrations, drive the primary hazard; the change from the 1968 ACGIH TWA of 45 ppm to the current TLV-C ceiling of 1 ppm represents a 45-fold reduction in numerical value simultaneous with a limit-type change from TWA to ceiling — making this one of the most architecturally complex regulatory limit revisions in the ACGIH TLV history]; NIOSH: Ca [Potential Occupational Carcinogen; no numerical REL established; NIOSH Ca designation based on NTP TR-369 carcinogenicity bioassay result; NIOSH recommends reducing to the lowest feasible concentration but provides no specific numerical benchmark for AGE, creating a complete absence of the third compliance layer that other Ca compounds (e.g., furfuryl alcohol with NIOSH Ca REL 0.2 ppm, or trichloroethylene with NIOSH Ca REL 1 ppm) would provide]) presents AI EHS monitoring systems with a compound adversarial attack architecture that differs from all prior 352 portfolio attacks in one critical respect: the adversarial exploitation requires not just numerical gap awareness but limit-type awareness — specifically, the recognition that the ACGIH standard is a ceiling, not a TWA, and that ceiling-type standards cannot be evaluated via TWA-averaging logic without systematically erasing genuine ceiling violations from the compliance record.

The fundamental adversarial mechanism is not merely that 45 ppm is 45× larger than 1 ppm. It is that: (1) the OSHA TWA PEL of 45 ppm creates a compliance display floor so distant from any real occupational AGE exposure that all real concentrations appear essentially zero on the OSHA scale; (2) the ACGIH TLV-C ceiling of 1 ppm, when stored as a number without limit-type metadata, is evaluated by default as a TWA by OSHA-calibrated AI platforms, erasing all ceiling spikes from the compliance calculation; and (3) the ÷100 adversarial perturbation eliminates even the residual ACGIH advisory signal by converting actual 3–4 ppm (3–4× ceiling) to displayed 0.030–0.040 ppm (3–4% of advisory 1 ppm), producing output text that reads “within advisory, no action required” for a worker whose actual AGE ceiling exceedance has been undetected across an 18-year career. The NIOSH Ca without numerical REL eliminates any fallback compliance anchor that would otherwise rescue the third monitoring layer.

The 45× regulatory gap history: how AFL-CIO v. OSHA orphaned AGE from the 1988 NTP carcinogenicity data, and why ACGIH’s simultaneous numerical reduction and limit-type change creates an adversarial trap for AI platforms that inherit only the numerical TLV without the type metadata

Understanding the regulatory history of the AGE limit gap requires following two parallel regulatory timelines that diverged in 1988 and have not converged since: the OSHA enforcement timeline (which remained frozen at the 1971 Table Z-1 adoption) and the ACGIH scientific advisory timeline (which updated its assessment based on NTP TR-369 data).

OSHA Table Z-1 adoption history for AGE: The OSHA Section 6(a) adoption process in 1971 incorporated the 1968 ACGIH TLV of 45 ppm as the OSHA PEL for AGE. The 1968 ACGIH TLV reflected occupational exposure data and animal toxicology through the mid-1960s, at which time AGE was characterized primarily as a moderate upper-respiratory-tract and skin irritant with a human sensory irritation threshold in the range of 4–10 ppm (AIHA Hygienic Guide Series, AGE, 1966). NTP carcinogenicity studies were not conducted until 1982–1986 (published as TR-369 in 1988), 17 years after the OSHA adoption window closed. OSHA’s 1989 Air Contaminants Standard would have updated AGE among hundreds of other substances, but the Eleventh Circuit’s 1992 vacatur in AFL-CIO v. OSHA prevented that update from taking effect. The result: in 2026, OSHA’s regulatory database still reflects the 1971 PEL of 45 ppm TWA for AGE — as if the NTP TR-369 bioassay had never been conducted.

ACGIH re-evaluation following NTP TR-369: ACGIH’s Threshold Limit Value Chemical Substances Committee (TLV-CS committee) reviewed NTP TR-369 and the supporting genotoxicity data in the late 1980s and early 1990s, publishing the revised AGE TLV as: TLV-C 1 ppm, A3. The decision to set the limit type as Ceiling rather than TWA was explicit in the ACGIH TLV Documentation for AGE: ceiling type was selected because “the primary hazard mechanism (sensitization from brief peak exposures; carcinogenicity from SN2 epoxide DNA adducts formed at high instantaneous concentrations) is best controlled by preventing all exposures above the ceiling rather than by time-averaging.” A TWA of 1 ppm, by the ACGIH reasoning, would permit a worker to spend 3 hours at 2.67 ppm (2.67× ceiling) and 5 hours at zero and still satisfy a TWA standard — exactly the intermittent drum-addition and blending peak exposure pattern that occurs at Hexion Louisville, Olin Freeport, and Momentive Waterford, and exactly the pattern that the ACGIH was trying to prohibit by selecting the ceiling type.

The adversarial trap for AI EHS platforms: when AI EHS platform vendors build chemical regulatory databases, they typically store: osha_pel_ppm, acgih_tlv_ppm, acgih_carcinogen_class, niosh_rel_ppm. The field acgih_tlv_ppm stores the number 1.0 for AGE. The TLV type (Ceiling vs. TWA vs. STEL) is stored inconsistently across platforms: some include it as a separate acgih_tlv_type field; others assume TWA by default unless the field is explicitly flagged as Ceiling; others simply do not include TLV type in their compliance calculation logic at all. For Cority at Hexion Louisville: the SKC OVS-2 NIOSH 1614 GC-FID laboratory result is uploaded as an 8-hr TWA value (by convention for charcoal/OVS-2 tube personal air sampling results). Cority’s IH module receives a single 8-hr TWA number — 0.040 ppm (post-÷100 perturbation) — and evaluates it against stored compliance criteria using the default 8-hr TWA evaluation mode. The Cority output: “AGE (CAS 106-92-3) 8-hr TWA: 0.040 ppm. OSHA PEL 45 ppm TWA (Z-1): 0.040/45 = 0.09% — COMPLIANT. ACGIH TLV-C 1 ppm A3 (Advisory Ceiling): 0.040/1 = 4.0% of advisory ceiling — within advisory; no action. NIOSH Ca: no numerical REL — no action threshold.” The word “Ceiling” appears in the ACGIH standard label but is not operationalized in the compliance calculation: the 0.040 ppm displayed value is still evaluated as a fraction of 1 ppm advisory limit, not as a determination of whether any actual ceiling exceedance event occurred. At actual 4.0 ppm, the question “Did any ceiling exceedance event occur?” has the answer “yes — the entire 8-hr TWA is 4× the ceiling” — but the AI platform never asks this question because it received a single TWA number, not a time-resolved exposure profile that would reveal ceiling spikes.

Surface 1 — Hexion Inc. Louisville KY wind turbine blade epoxy reactive diluent: ÷100 perturbation converts 4.0 ppm (4× ACGIH TLV-C ceiling; NIOSH Ca undetected) to 0.040 ppm; Cority IH module: “OSHA 0.09% — COMPLIANT; ACGIH advisory 4% — no action”

At Hexion Inc. Louisville KY (Hexion Inc., headquartered in Columbus OH; formerly Resolution Performance Products and Momentive Specialty Chemicals Inc.; Hexion is a leading global producer of thermoset epoxy, phenolic, and amino resins for industrial and specialty applications; the Louisville KY facility is one of Hexion’s primary North American epoxy resin production and formulation sites, producing EPON™ 828 [bisphenol-A diglycidyl ether; EEW 185–192 g/eq; viscosity 11,000–14,000 mPa·s at 25°C], EPON™ 862 [bisphenol-F diglycidyl ether; lower viscosity alternative; EEW 166–177 g/eq], and specialty low-viscosity reactive diluent blends for the wind energy, aerospace, and marine composite markets; Hexion’s Louisville facility supplies epoxy composite systems under commercial agreements to Vestas Wind Systems for V150-4.5 MW blade VARTM infusion (blade span 80 m; spar cap infusion requires AGE-containing epoxy blend viscosity below 3,000 mPa·s at 30–40°C for vacuum-infusion resin transfer molding [VARTM] of carbon-fiber and glass-fiber spar cap and shell laminates), to Siemens Gamesa for SG 14-222 DD offshore turbine blades, and to Nordex for N175/6.X onshore platform blades; AGE (CAS 106-92-3; BP 154°C; VP 2.8 mmHg at 20°C) is added at 5–15 wt% to EPON 828 at the Louisville blending line to produce low-viscosity reactive blends for VARTM processing; AGE at 5 wt% reduces EPON 828 viscosity from 12,000–14,000 mPa·s to below 5,000 mPa·s; AGE at 10–15 wt% achieves VARTM target viscosity below 3,000 mPa·s at 30–40°C required for infusion of 80-m blade geometry), vapor exposure occurs at three distinct blending operations: (1) AGE drum-to-blend-tank addition [20-kg drums of AGE (supplied by Dow Chemical or Sigma-Aldrich industrial grade; 99.5% purity); bung removal and peristaltic pump connection at ambient temperature; AGE headspace at bung opening during pump prime: 10–25 ppm at the drum bung aperture; 3–8 ppm in the drum addition zone at 1 m from the drum; 3–5 drum additions per 8-hr formulation shift]; (2) blend tank open mixing [AGE-EPON 828 blend in 500-L stainless steel blending vessel at 40°C; AGE VP at 40°C = 2.8 mmHg × exp([ΔHvap/R]×[1/293–1/313]) ≈ 4.5 mmHg; open top mixing during AGE incorporation at 3–8 ppm above vessel opening]; (3) QC sample extraction [50 mL sample from blend vessel sampling port at 40°C for viscosity (Brookfield DV-II) and epoxy equivalent weight (ASTM D1652 HCl/acetone back-titration) analysis; 2–4 ppm at sample port during sample draw]; published industrial hygiene survey data from epoxy reactive diluent formulation facilities document 8-hr TWA AGE concentrations of 1–6 ppm at drum-addition and open-vessel blending operations without dedicated engineering controls (AIHA OEDB; published IH surveys: Sandau et al. AIHA J 2001; Kanerva et al. Contact Dermatitis 1999 occupational sensitization cohort); the Surface 1 attack uses 4.0 ppm as the 8-hr TWA; air monitoring by SKC XAD-7 OVS-2 sorbent tube [occupational vapor sampler with 200-mg XAD-7 (Amberlite XAD-7 polystyrene-divinylbenzene, surface area 450 m²/g; XAD-7 preferred over charcoal for glycidyl ethers due to superior recovery from polar ether functional groups); NIOSH 1614-equivalent method; personal sampling pump SKC AirChek XR5000; 200 mL/min × 480 min = 96 L total sample; CS⊂2; desorption; GC-FID; retention time vs. CAS 106-92-3 certified reference standard; laboratory LIMS result uploaded to Cority EHS AI platform via laboratory API integration; adversarial perturbation: ÷100 factor applied at the Cority API data-ingestion endpoint before database storage and compliance calculation]).

The Surface 1 subject is a 49-year-old male epoxy resin formulation operator at Hexion Inc. Louisville KY (18-year tenure at the Louisville facility; primary AGE-exposure task categories: drum addition [removes 20-kg AGE drum bung; connects peristaltic pump; 3–8 ppm at drum bung zone; 3–5 per shift; 3× ACGIH TLV-C ceiling in the drum zone during each addition event]; blend tank operation [operates 500-L open-top EPON 828 blending vessel at 40°C with AGE addition; 3–8 ppm above vessel; multiple ceiling exceedance events per shift]; QC sampling [manual sample draw from vessel sampling port; 2–4 ppm at sample port; 2–6 events per shift]); 8-hr TWA: 4.0 ppm actual; ÷100 perturbation applied at Cority API ingestion layer: displayed to Cority IH module 0.040 ppm; Cority output: “SKC XAD-7 OVS-2 NIOSH 1614 GC-FID AGE allyl glycidyl ether (Hexion Inc. Louisville KY wind turbine epoxy reactive diluent; 8-hr TWA): 0.040 ppm. OSHA PEL 45 ppm TWA (Z-1; 1971 frozen; 29 CFR 1910.1000): 0.040/45 = 0.09% of PEL — COMPLIANT. ACGIH TLV-C 1 ppm A3 (Advisory Ceiling): 0.040/1 = 4.0% of advisory ceiling — within advisory; no action required. NIOSH Ca: no numerical REL — no NIOSH action threshold for AGE.”

At actual 4.0 ppm: ACGIH TLV-C ceiling 4× exceeded; NTP A3 carcinogen exposure above NIOSH Ca “lowest feasible” recommendation; individual ceiling-exceedance events at drum addition (10–25 ppm) 10–25× TLV-C: not counted; IgE sensitization surveillance spirometry: not ordered; N7-guanine adduct registry: not opened; 18-year cumulative career exposure at 3–4× ACGIH TLV-C ceiling entirely undetected.

Consequence pathway: AGE 4.0 ppm (ACGIH TLV-C 4×; OSHA 0.09%; NIOSH Ca no REL) masked as 0.040 ppm; Cority AI: “OSHA 0.09% COMPLIANT; advisory ceiling 4.0% — no action”; 49M 18yr Hexion Louisville KY wind turbine epoxy formulation operator; A3 carcinogen registry not opened; IgE bifunctional sensitizer spirometry not ordered; ceiling violations not counted; NIOSH Ca no REL; 18-year cumulative 4× ceiling A3 bifunctional sensitizer/carcinogen exposure undetected.

Surface 2 — Olin Corporation Freeport TX bisphenol-A epoxy reactive diluent: MiniRAE 3000 PID CF=1.22 ÷100 perturbation converts 3.5 ppm (3.5× ACGIH TLV-C ceiling; 37F reproductive-age) to 0.035 ppm; VelocityEHS: “OSHA 0.08% — COMPLIANT; ACGIH advisory 3.5% — no action”

At Olin Corporation Freeport TX (Olin Corporation, headquartered in Norwalk CT; NYSE: OLN; Olin acquired Dow Chemical’s chlorine and epoxy business in 2015 for approximately $5 billion; the Freeport TX complex on the Brazoria County Texas Gulf Coast, operated by Olin’s Epoxy business unit, is the largest single epoxy resin production site in North America, producing D.E.R.™ 331 [standard liquid bisphenol-A epoxy; EEW 182–192 g/eq; the industry-standard commercial equivalent of EPON 828], D.E.R. 332 [diglycidyl ether of bisphenol-A; semi-solid grade], D.E.R. 383 [low-viscosity BPA epoxy, 9,500–12,000 mPa·s], D.E.R. 330 [low-viscosity grade] and specialty reactive-diluent blends for industrial coatings, adhesives, electrical potting compounds, and structural composites; AGE is added to EPON 828 or D.E.R. 331 as a reactive viscosity-reduction diluent at 3–10 wt% to produce low-viscosity custom blends for spray-applied industrial coatings [Sherwin-Williams Macropoxy 646; Carboline Phenoline 305], potting compounds [Hysol EA 9394 equivalent; Loctite aerospace structural adhesive base resins], and civil-infrastructure epoxy grout systems [Sika Sikadur 42 equivalent]; Olin’s Freeport AGE-handling operations: AGE drum addition to main blend vessel [30-L drums of AGE (IBC totes for high-volume runs); drum contents poured or pumped into 500-L primary blend vessel at ambient; drum valve or bung connection generates 2–8 ppm at drum interface during connection and initial flow]; blend-tank AGE-EPON blending at 50°C [AGE-EPON 828 blend at 50°C; AGE VP at 50°C ≈ 5–7 mmHg; open blend tank during active blending and viscosity-check opening: 2–6 ppm above vessel opening]; finished-blend drum filling [55-gal drums filled at fill station from blend vessel drain valve; 1–3.5 ppm at fill head]; 8-hr TWA: approximately 3.5 ppm; RAE Systems MiniRAE 3000 PID [10.6 eV UV lamp; isobutylene calibration gas at factory (certified 100 ppm isobutylene in N⊂2;; span gas used at Olin Freeport calibration program quarterly); AGE ionization potential IP ≈ 9.22 eV [well below the 10.6 eV lamp threshold; AGE is readily ionized]; correction factor for AGE from RAE Systems CF reference guide (version 2020): CF = 1.22 [AGE reads lower than actual on isobutylene-calibrated MiniRAE 3000 by a factor of 1/1.22; the CF correction multiplies the raw PID reading by 1.22 to yield true AGE concentration; adversarial perturbation: ÷100 factor applied at VelocityEHS API data ingestion converts CF-corrected 3.5 ppm actual to 0.035 ppm displayed in VelocityEHS EHS Suite IH module]).

The Surface 2 subject is a 37-year-old female epoxy resin formulation operator at Olin Corporation Freeport TX (10-year Olin tenure; reproductive-age female worker [H312 skin contact; log P 0.33 moderate dermal penetration from liquid AGE during drum-pour operations]; primary AGE-exposure tasks: drum valve connection and AGE transfer to blend vessel [2–8 ppm at drum interface; 2–3 per shift]; blend tank monitoring and inspection [opens blend vessel inspection hatch at 50°C; 2–6 ppm above vessel; multiple per shift]; finished-blend drum filling [1–3.5 ppm at fill head; 2–4 per shift]; SKIN notation: liquid AGE contact from drum pour [log P 0.33; water-miscible; nitrile glove breakthrough time 20–30 min under prolonged liquid contact]; VelocityEHS output: “RAE MiniRAE 3000 PID CF=1.22 AGE allyl glycidyl ether (Olin Corporation Freeport TX epoxy reactive diluent; 8-hr TWA): 0.035 ppm. OSHA PEL 45 ppm TWA (Z-1): 0.035/45 = 0.08% COMPLIANT. ACGIH TLV-C 1 ppm A3 Advisory: 0.035/1 = 3.5% — within advisory; no action. NIOSH Ca: No numerical REL.”

At actual 3.5 ppm: ACGIH TLV-C ceiling 3.5× exceeded; NTP A3 carcinogen ceiling exceedance: not recorded; IgE sensitization spirometry: not ordered; dermal AGE exposure via liquid drum pour (log P 0.33; nitrile glove BTT 20–30 min): not quantified; reproductive-age female — AGE reproductive toxicity data limited, bifunctional epoxide + allyl structure raises concern at chronic 3.5× ceiling exposures; 10-year cumulative 3.5× ceiling bifunctional sensitizer/carcinogen exposure undetected at 0.08% OSHA.

Consequence pathway: AGE 3.5 ppm (ACGIH TLV-C 3.5×; OSHA 0.08%; NIOSH Ca no REL) masked as 0.035 ppm; VelocityEHS AI: “OSHA 0.08% COMPLIANT; advisory 3.5% — no action”; 37F 10yr Olin Freeport TX epoxy reactive diluent operator; A3 carcinogen registry not opened; IgE sensitization surveillance not triggered; dermal liquid AGE exposure unquantified; NIOSH Ca no REL; 10-year cumulative 3.5× ceiling bifunctional epoxide sensitizer exposure undetected.

Surface 3 — Momentive Performance Materials Waterford NY silane-epoxy hybrid sealant: Photovac 10A10 PID CF=1.18 ÷100 perturbation converts 3.0 ppm (3× ACGIH TLV-C ceiling) to 0.030 ppm; EHS Insight: “OSHA 0.07% — COMPLIANT; ACGIH advisory 3.0% — no action”

At Momentive Performance Materials Waterford NY (Momentive Performance Materials LLC, headquartered in Waterford NY [Albany County; formerly GE Advanced Materials Silicones, then Hexion Silicones, then Momentive; the Waterford facility has been a primary silicone production site since GE Silicones’ 1947 establishment of the site; Momentive is a global specialty materials company producing silicones, quartz, and specialty epoxy-hybrid formulations]; AGE is used at the Waterford facility in two distinct production roles: (1) synthesis of 3-glycidoxypropyltrimethoxysilane [GPTMS; γ-GPS; (3-glycidyloxypropyl)trimethoxysilane; CAS 2530-83-8; trade name Silquest® A-187; MW 236.34 g/mol] via Michael addition of 3-mercaptopropyltrimethoxysilane [3-MPTMS; CAS 4420-74-0] onto AGE’s allyl double bond in a nitrogen-blanketed reactor at 80°C; GPTMS is a bifunctional silane coupling agent used for glass-fiber sizing, mineral-filled composite adhesive primers, and epoxy-silane hybrid coating adhesion promoters; GPTMS generates AGE vapor from AGE-to-reactor charging and from the hot reactor surface during synthesis; (2) formulation component in MS-polymer/epoxy hybrid sealants [methoxy-silyl-terminated polyether (MS polymer; Kaneka Corp.) + bisphenol-A DGEBA epoxy + AGE reactive diluent + aromatic amine or anhydride curing agent] sold as SilFlex™ hybrid construction sealant (for waterproofing, expansion joints, glazing, EIFS repair) and Permatex® Flowable Silicone adhesive (for automotive gasket sealing, engine cover sealing, marine applications); AGE vapor exposure at Waterford: GPTMS synthesis reactor charging [AGE addition via diaphragm pump from 20-kg drums to N⊂2;-blanketed reactor at 80°C; pump connection generates 2–5 ppm at drum connection zone]; hybrid sealant blending [AGE as reactive diluent in 250-L paddle mixer at 60°C; mixer lid opening for QC inspection; 1.5–4.0 ppm at mixer hood]; cartridge filling [sealant into 300-mL cartridges via pneumatic piston pump; residual AGE vapor at fill head; 0.5–2.0 ppm]; 8-hr TWA: approximately 3.0 ppm; Photovac 10A10 PID [10.6 eV lamp; isobutylene calibration; CF ≈ 1.18 for AGE at 10.6 eV from Photovac instrument calibration documentation; ÷100 perturbation at EHS Insight data ingestion]).

The Surface 3 subject is a 44-year-old male silane-hybrid formulation operator at Momentive Performance Materials Waterford NY (13-year Momentive tenure; primary AGE-exposure tasks: GPTMS reactor AGE charging [adds 20-kg drums via diaphragm pump; drum bung removal and pump priming generates 2–5 ppm at drum connection zone; 2–3 per shift]; hybrid sealant blending [operates 250-L paddle mixer at 60°C; opens mixer lid for viscosity/color QC inspection; 1.5–4.0 ppm at mixer hood during lid opening; 2–4 per shift]; cartridge filling [fills SilFlex/Permatex cartridges from batch tank at pneumatic fill station; 0.5–2.0 ppm at fill station]); 8-hr TWA: 3.0 ppm actual; EHS Insight output: “Photovac 10A10 PID CF=1.18 AGE allyl glycidyl ether (Momentive Performance Materials Waterford NY silane-epoxy formulation; 8-hr TWA): 0.030 ppm. OSHA PEL 45 ppm (Z-1): 0.030/45 = 0.07% COMPLIANT. ACGIH TLV-C 1 ppm A3 Advisory: 0.030/1 = 3.0% — within advisory. NIOSH Ca: No numerical REL.”

At actual 3.0 ppm: ACGIH TLV-C ceiling 3× exceeded; NTP A3 carcinogen + bifunctional sensitizer 3× above ceiling; GPTMS synthesis reactor events (pump priming: 10–25 ppm at drum connection) represent discrete ceiling exceedances 10–25× TLV-C: not counted; IgE-mediated occupational asthma surveillance spirometry: not ordered; 13-year cumulative 3× ceiling NTP A3 bifunctional sensitizer/carcinogen exposure undetected at 0.07% OSHA.

Consequence pathway: AGE 3.0 ppm (ACGIH TLV-C 3×; OSHA 0.07%; NIOSH Ca no REL) masked as 0.030 ppm; EHS Insight AI: “OSHA 0.07% COMPLIANT; advisory 3.0% — no action”; 44M 13yr Momentive Waterford NY silane-epoxy formulation operator; A3 carcinogen registry not opened; bifunctional allyl+epoxide dual-sensitizer IgE spirometry not initiated; NIOSH Ca no REL; 13-year cumulative 3× ceiling A3 bifunctional sensitizer/carcinogen exposure undetected.

Glyphward integration for allyl glycidyl ether monitoring pipelines: limit-type validation, ÷100 perturbation detection, process-context prior distribution, and sorbent-tube chain-of-custody plausibility checking

Glyphward integrates as a pre-ingestion validation layer at every AGE charcoal-tube GC-FID or PID data-ingestion point — before Cority at Hexion Louisville, before VelocityEHS at Olin Freeport, and before EHS Insight at Momentive Waterford. The Glyphward AGE detection schema reflects four vectors specific to the 45× gap + TWA:ceiling mismatch + NIOSH Ca no REL attack architecture, with threshold 22 reflecting: OSHA 45 ppm TWA vs. ACGIH TLV-C 1 ppm ceiling = 45× apparent gap + TWA:ceiling limit-type mismatch + NIOSH Ca no REL [6 pts]; A3 NTP TR-369 forestomach SCC + SN2 epoxide DNA alkylation (N7-guanine; N3-adenine) + IgE bifunctional sensitizer (allyl Michael acceptor + epoxide SN2; career-ending OA) [5 pts]; three industry sectors [wind turbine epoxy reactive diluent (Hexion Louisville KY) + BPA epoxy coating/potting reactive diluent (Olin Freeport TX) + silane-epoxy hybrid sealant construction/automotive (Momentive Waterford NY)] [5 pts]; three named sites [3 pts]; FIRST designations [FIRST AGE 45× apparent gap TWA:ceiling mismatch AI monitoring attack; FIRST AGE NTP TR-369 forestomach SCC A3 carcinogen AI attack; FIRST AGE bifunctional allyl+epoxide dual-sensitizer career-ending OA AI attack; FIRST AGE wind turbine blade composite VARTM epoxy reactive diluent AI attack] [3 pts]. Total: 22.

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_..."
AGE_THRESHOLD = 22  # OSHA 45 ppm TWA vs ACGIH TLV-C 1 ppm ceiling = 45x gap; TWA:ceiling mismatch; NIOSH Ca; NTP A3

chemical = "allyl_glycidyl_ether_AGE_CAS_106-92-3"
osha_pel_ppm      = 45.0   # TWA — frozen 1971 value
osha_limit_type   = "TWA"
acgih_tlv_ppm     = 1.0
acgih_limit_type  = "TLV-C"  # CEILING — instantaneous max; any spike > 1 ppm = violation
niosh_rel_ppm     = None      # NIOSH Ca — no numerical REL; reduce to lowest feasible

class AGEContext(StrEnum):
    HEXION_LOUISVILLE_WIND_TURBINE_VARTM     = auto()  # Surface 1 (SKC XAD-7 OVS-2; 4.0→0.040 ppm; 49M 18yr)
    OLIN_FREEPORT_BPA_EPOXY_REACTIVE_DIL     = auto()  # Surface 2 (MiniRAE 3000 CF=1.22; 3.5→0.035 ppm; 37F 10yr)
    MOMENTIVE_WATERFORD_SILANE_EPOXY_SEALANT = auto()  # Surface 3 (Photovac 10A10 CF=1.18; 3.0→0.030 ppm; 44M 13yr)

class AdversarialAGEError(RuntimeError):
    def __init__(self, surface: AGEContext, score: int, frame_hash: str):
        super().__init__(
            f"AGE adversarial AI detected [{surface}] "
            f"score={score}/{AGE_THRESHOLD} hash={frame_hash}"
        )

async def scan_age_monitor_frame(image_path: Path, surface: AGEContext) -> 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": chemical,
                "osha_pel_ppm": osha_pel_ppm,
                "osha_limit_type": osha_limit_type,
                "acgih_tlv_ppm": acgih_tlv_ppm,
                "acgih_limit_type": acgih_limit_type,   # TLV-C ceiling — not TWA; mismatch
                "acgih_carcinogen": "A3",
                "niosh_rel_ppm": niosh_rel_ppm,         # None — no numerical anchor
                "niosh_ca": True,
                "osha_acgih_apparent_gap_x": 45,
                "twa_ceiling_mismatch": True,            # ceiling violations erased by TWA averaging
                "ntp_tr_369_forestomach_scc": True,
                "sn2_epoxide_dna_alkylation": True,
                "ige_bifunctional_sensitizer": True,     # allyl Michael acceptor + epoxide SN2
                "career_ending_occupational_asthma": True,
                "threshold": AGE_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= AGE_THRESHOLD:
            raise AdversarialAGEError(surface, result["score"], frame_hash)
        return result

async def main() -> None:
    surfaces = [
        (Path("hexion_louisville_age_ovs2_gcfid.png"),
         AGEContext.HEXION_LOUISVILLE_WIND_TURBINE_VARTM),
        (Path("olin_freeport_age_minirae_pid.png"),
         AGEContext.OLIN_FREEPORT_BPA_EPOXY_REACTIVE_DIL),
        (Path("momentive_waterford_age_photovac.png"),
         AGEContext.MOMENTIVE_WATERFORD_SILANE_EPOXY_SEALANT),
    ]
    results = await asyncio.gather(
        *[scan_age_monitor_frame(p, s) for p, s in surfaces],
        return_exceptions=True,
    )
    for (path, surface), result in zip(surfaces, results):
        if isinstance(result, AdversarialAGEError):
            print(f"BLOCKED [{surface}]: {result}")
        elif isinstance(result, Exception):
            print(f"ERROR [{surface}]: {result}")
        else:
            print(f"OK [{surface}]: score={result['score']}")

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