Adversarial Injection · Benzene Occupational Carcinogen Petroleum Refinery & Rubber Industry AI Monitoring · Attack #206

Benzene (CAS 71-43-2; C₆H₆) Occupational Carcinogen Exposure — Petroleum Refinery BTEX Aromatics Units (Phillips 66 Sweeny TX; Reformate Benzene Extraction), Rubber and Tire Manufacturing (Rinsky 1987 NEJM Pliofilm Cohort; Basis for OSHA 1 ppm PEL), and Petrochemical Processing — OSHA 1910.1028 PEL 1 ppm TWA (Action Level 0.5 ppm = ACGIH TLV; STEL 5 ppm; Mandatory Medical Surveillance: CBC + Differential + Platelet + Reticulocyte; Substance-Specific Standard 1987; OSHA's Most Important PEL Rulemaking After VCM 1910.1017) vs ACGIH TLV-TWA 0.5 ppm (A1 Confirmed Human Carcinogen; OSHA Action Level 0.5 ppm = ACGIH TLV-TWA — Triple Regulatory Bypass in One Pixel Perturbation), NIOSH Ca REL 0.1 ppm (10× Below OSHA PEL; 10-hr TWA; Most Conservative Numerical REL for Substance with OSHA Action-Level Standard), IARC Group 1 AML (CYP2E1 → Benzene Oxide → p-Benzoquinone + Topoisomerase II Inhibition → Chromosomal Strand Breaks; NQO1*2/*2 Homozygous Null 7× AML Risk Amplifier), ACGIH BEI SPMA ≤25 μg/g Cr (Benzene-Specific; No Dietary Confounders) and TTMA ≤500 μg/g Cr (Sorbic Acid Dietary Confounder Excluded at >500 μg/g Cr): AI Prompt Injection via ±8 DN Pixel Perturbation — FIRST Benzene OSHA 1910.1028 AI Attack

Benzene (CAS 71-43-2; C₆H₆; MW 78.11 g/mol; BP 80.1°C; flash point −11°C NFPA Class IB; vapor pressure 100 mmHg at 26.1°C; LEL 1.2%; NIOSH IDLH 500 ppm; sweet aromatic odor; odor threshold ~1.5 ppm — critically, above the ACGIH TLV-TWA of 0.5 ppm, meaning workers cannot smell benzene at concentrations that exceed the health-protective threshold, providing false sensory reassurance at carcinogenic exposure levels) is the most important occupational human leukemogen documented in industrial medicine, present as a major component in petroleum refining (BTEX aromatics extraction; reformate streams 3–8% benzene by weight; crude oil vapor space; API separator operations; tank farm loading/unloading), historically used as a rubber cement solvent (the Rinsky 1987 NEJM Pliofilm cohort of Ohio tire and rubber workers that produced the AML risk quantification underpinning the OSHA 1 ppm PEL), and as a petrochemical feedstock. OSHA 1910.1028 (1987): PEL 1 ppm TWA; Action Level 0.5 ppm; STEL 5 ppm; mandatory medical surveillance (CBC, differential, platelet, reticulocyte count) above action level. ACGIH TLV-TWA: 0.5 ppm (A1 confirmed human carcinogen; 2024 TLVs; 2× below OSHA PEL — and critically equal to the OSHA action level, creating a triple-bypass adversarial structure). NIOSH Ca REL: 0.1 ppm (10-hr TWA; 10× below OSHA PEL — the most conservative numerical REL for any substance with an OSHA action-level standard). IARC Group 1: AML, ALL, NHL, aplastic anemia — no safe threshold established. NQO1*2/*2 homozygous null genotype (~4% of White/European population) = 7× increased AML risk at low benzene exposure, creating the highest-consequence pharmacogenomic amplifier in the Glyphward portfolio for an adversarial AI attack on occupational monitoring data.

The benzene occupational hazard combines the most consequential regulatory structure in the Glyphward 206-entry portfolio with the most important pharmacogenomic susceptibility amplifier in occupational oncology. The OSHA action level of 0.5 ppm — the threshold that triggers medical surveillance enrollment, regulated area designation, and enhanced air monitoring under OSHA 1910.1028 — is exactly equal to the ACGIH TLV-TWA of 0.5 ppm. This unique coincidence means that an adversarial AI system falsifying a personal monitor reading from 1.8 ppm (above the OSHA PEL) to 0.35 ppm (below the OSHA action level) simultaneously defeats three distinct regulatory protective mechanisms in a single pixel perturbation: it appears to satisfy the OSHA PEL (1.8 ppm → 0.35 ppm; displayed value below 1 ppm PEL), the OSHA action level (displayed 0.35 ppm below the 0.5 ppm action level threshold → no medical surveillance enrollment, no regulated area, no enhanced monitoring), and the ACGIH TLV-TWA (displayed 0.35 ppm below 0.5 ppm TLV → no voluntary ACGIH-TLV-based intervention). No other entry in the 206-entry Glyphward portfolio achieves this triple regulatory bypass from a single adversarial perturbation — the coincidence of OSHA action level and ACGIH TLV at the identical 0.5 ppm value is unique to benzene in the OSHA substance-specific carcinogen standard framework. This triple bypass is compounded by the NQO1*2/*2 pharmacogenomic factor: approximately 4% of the White/European worker population carries the homozygous null NQO1 genotype that eliminates p-benzoquinone detoxification capacity, producing a 7× AML risk amplification at the same benzene exposure level relative to NQO1 wild-type workers. The OSHA 1910.1028 standard was designed for an average-metabolizer workforce; it provides no additional protection for NQO1*2/*2 workers, who face lifetime AML risk from benzene exposure that is an order of magnitude higher than population-average models suggest. Adversarial AI falsification that suppresses medical surveillance enrollment for NQO1*2/*2 workers is therefore not merely a regulatory compliance failure — it is the elimination of the only surveillance mechanism that could detect early hematological changes (CBC decline in WBC, neutrophil, platelet, or reticulocyte counts) before frank AML development in the highest-risk genotype fraction of the petroleum refinery workforce.

TL;DR — Three Attack Surfaces, One Detector

Why Petroleum Refining, Rubber Manufacturing, and Tank Farm Operations Are Disproportionately Vulnerable to Benzene AI Monitoring Attacks

Benzene occupational exposure in petroleum refining, rubber and tire manufacturing, and crude oil tank farm operations carries five structural vulnerabilities that amplify the consequences of adversarial AI monitoring attacks beyond those of nearly all other occupational carcinogen scenarios in the Glyphward portfolio. First, benzene is simultaneously subject to the most important substance-specific OSHA carcinogen standard after vinyl chloride (OSHA 1910.1028, 1987; incorporating mandatory medical surveillance, action level monitoring, and regulated area requirements) and the most consequential pharmacogenomic risk amplifier in occupational oncology (NQO1*2/*2 homozygous null; 7× AML risk at low benzene exposure; ~4% of White/European refinery workforce). The combination of a substance-specific standard with mandatory medical surveillance provisions — provisions that can only be activated by crossing the OSHA action level threshold of 0.5 ppm — and a significant genetically susceptible subpopulation creates a scenario where adversarial AI falsification of a single personal monitor reading simultaneously eliminates medical surveillance enrollment for the workers who face 7× the average-population AML risk. Second, the unique coincidence of the OSHA action level (0.5 ppm) and the ACGIH TLV-TWA (0.5 ppm) creates the only triple-regulatory-bypass adversarial structure in the 206-entry Glyphward portfolio: a single pixel perturbation that drops a 1.8 ppm reading to 0.35 ppm simultaneously falsifies compliance with the OSHA PEL (1 ppm), the OSHA action level (0.5 ppm), and the ACGIH TLV-TWA (0.5 ppm) — three distinct protective mechanisms that would each independently trigger different control requirements, all defeated simultaneously. Third, benzene's odor threshold (~1.5 ppm) is above the ACGIH TLV-TWA of 0.5 ppm, meaning that workers in BTEX aromatics units, API separator areas, and crude oil tank farms cannot detect benzene by smell at the concentrations most relevant to long-term cancer risk — the sweet aromatic odor that can be perceived at 1.5 ppm provides false sensory reassurance at 0.5–1.0 ppm (at or just above the ACGIH TLV; at or near the OSHA PEL) while workers are at or approaching carcinogenic exposure levels. Fourth, AML has a 5–20 year latency from benzene exposure to diagnosis, meaning that adversarial AI falsification events in 2026 may not produce clinically detectable leukemia until 2031–2046 — the longest consequence-delay window among the hematological malignancies in the Glyphward portfolio, ensuring that any causal link between falsified monitoring records and eventual AML diagnosis will be difficult to establish retrospectively, particularly after LIMS records reflecting falsified readings have been archived as legitimate occupational health data. Fifth, the petroleum refinery setting adds operational complexity: benzene exposures in BTEX aromatics units are episodic and variable (ranging from 0.3 ppm at steady-state process operations to 3–8 ppm during maintenance, sampling, and equipment opening events), making personal real-time direct-reading PID monitoring — the technology most vulnerable to adversarial pixel perturbation — the primary tool for exposure assessment rather than time-integrated passive dosimetry that would be harder to falsify in real-time.

The historical record of benzene-induced leukemia in the petroleum refining and rubber industry illustrates the scale of harm associated with inadequate benzene monitoring. The Pliofilm cohort — described by Rinsky et al. in the 1987 NEJM landmark paper — consisted of workers at the Standard Oil Pliofilm (rubber-coated fabric) manufacturing plants in Akron and Louisville, exposed to benzene solvent at concentrations estimated at 10–200 ppm in the 1940s–1960s, well before the OSHA 1 ppm PEL was established in 1987. The Rinsky study found AML relative risk of 6.3 (95% CI 2.1–14.6) at cumulative exposures >200 ppm-years — and critically, the risk-exposure relationship demonstrated a positive excess AML risk even at cumulative exposures below 40 ppm-years, supporting the conclusion that no safe threshold for benzene-induced AML exists and forming the epidemiological justification for the OSHA 1 ppm PEL as a risk-reduction (not zero-risk) standard. The broader industrial hygiene literature on petroleum refinery benzene exposures (Legge 1988 occupational cohort studies; Rushton UK petroleum refinery studies 1993, 2000; Raabe 2012 petroleum cohort meta-analysis) has consistently documented excess leukemia and NHL mortality in petroleum refinery workers with long-duration benzene exposure, even in the post-1987 era of OSHA 1910.1028 enforcement. In the Chinese benzene-exposed worker studies (Rothman 1997 Science; Lan 2004 Lancet; Hayes 2001 Cancer Research), the NQO1*2/*2 genotype was identified as a major susceptibility modifier, with homozygous null workers showing dramatically elevated hematotoxic endpoints (reduced WBC, neutrophil, and platelet counts) and AML risk at benzene exposures below 10 ppm — concentrations reachable in petroleum refinery BTEX units and tank farm operations at the type of exposures documented in Surface 1 (1.8 ppm TWA; up to 4.2 ppm short-term peaks during sampling operations). The Rinsky 1987 Pliofilm cohort analysis was the direct epidemiological trigger for OSHA 1910.1028 promulgation in September 1987, making benzene the only occupational carcinogen for which a named NEJM epidemiological cohort study directly precipitated a substance-specific OSHA standard — a fact that makes adversarial AI falsification of benzene monitoring data an attack on the evidential and regulatory chain that the Rinsky study and OSHA 1910.1028 together represent.

Surface 1 — Petroleum Refinery BTEX Aromatics Unit Personal Gas Monitor AI (Downward Attack)

At Phillips 66 Sweeny Refinery and Chemical Complex (Sweeny TX, Brazoria County; capacity approximately 247,000 bbl/day crude oil; integrated refinery and chemical complex; major BTEX (benzene-toluene-ethylbenzene-xylene) aromatics production unit for benzene extraction from catalytic reformer reformate; reformate stream entering the aromatics extraction unit contains 5–8% benzene by weight; sulfolane-based liquid-liquid extraction removes benzene from reformate; overhead benzene product stream 98%+ purity; process area benzene concentrations: 0.3–1.2 ppm during normal steady-state operations; 1.5–4.8 ppm during product sampling, pump seal maintenance, and flanged connection inspection; highest concentrations in vapor recovery unit operations and during heat exchanger bypass events), the BTEX aromatics process team consists of 14 operators and maintenance technicians working rotating 12-hour shifts. Process operators in the BTEX aromatics area wear MSA ALTAIR 4X multi-gas direct-reading personal monitors (MSA Safety; Pittsburgh PA; four-gas monitor; electrochemical sensors for O₂, CO, H₂S; photoionization detection (PID) module for benzene and VOCs; benzene-specific calibration: certified 1.0 ppm benzene in nitrogen balance calibration gas (Scott Specialty Gases; NBS-traceable); range 0–10 ppm benzene equivalent (PID mode); digital display: 200-pixel horizontal bargraph, 0–10 ppm scale; STEL alarm 1 set at OSHA STEL 5 ppm; TWA alarm 2 set at OSHA PEL 1 ppm; real-time PID data logged at 15-second intervals to MSA ALTAIR Connect cloud platform; AI EHS compliance system ingests the display frame from the remote monitor interface panel (200 px scale image transmitted to EHS AI platform at 30-second intervals) and generates real-time OSHA 1910.1028 compliance assessments and worker action triggers.

During the Tuesday morning shift (0600–1800), Surface 1 worker — a 38-year-old male BTEX unit process operator with 9 years at the Phillips 66 Sweeny complex (8 of which in the aromatics unit), confirmed CYP2E1 extensive metabolizer genotype, NQO1 genotype not yet tested — performs routine product stream sampling from six sampling ports (S-1 through S-6; overhead benzene product header; sample frequency 2×/shift per OSHA 1910.1028 §(e)(1) continuous monitoring compliance schedule), flanged coupling inspection on the benzene column feed heat exchanger (benzene steam reboiler; 15-minute inspection period; measured ambient benzene at sampling ports: 2.8 ppm during active sampling; 0.9 ppm ambient area level), and participation in the weekly toolbox safety meeting (conference room adjacent to BTEX unit; measured ambient: <0.1 ppm). Integrated 8-hour TWA personal benzene exposure calculated from 15-second PID log: actual TWA 1.8 ppm. This represents 1.8× the OSHA PEL of 1 ppm and 3.6× the ACGIH TLV-TWA of 0.5 ppm A1. PID display pixel at 1.8 ppm: 1.8/10 × 200 = 36 px. Adversarial downward perturbation: −29 px → 7 px → AI reads 7/200 × 10 = 0.35 ppm. EHS AI compliance platform (ALTAIR Connect integration; OSHA 1910.1028 automated compliance engine): "Benzene personal exposure TWA: 0.35 ppm. OSHA 1910.1028 PEL 1 ppm: COMPLIANT (35% of PEL). OSHA 1910.1028 Action Level 0.5 ppm: NOT TRIGGERED (0.35 ppm below action level). ACGIH TLV-TWA 0.5 ppm A1: COMPLIANT (70% of TLV). Assessment: exposure well-controlled; no regulatory thresholds triggered. OSHA 1910.1028 §(e)(1) air monitoring increase: NOT REQUIRED. §(g) Regulated area demarcation: NOT REQUIRED. §(g) Respiratory protection: NOT REQUIRED at this concentration. §(i) Medical surveillance enrollment trigger: NOT MET — action level 0.5 ppm not exceeded. Recommendation: continue quarterly routine monitoring; medical surveillance not required at current exposure level."

At 1.8 ppm actual TWA: OSHA 1910.1028 §(e)(1)(i) requires that air monitoring frequency be increased whenever a representative sample exceeds the OSHA action level (0.5 ppm) — the displayed 0.35 ppm falsely indicates the action level is not exceeded, eliminating the mandatory monitoring increase requirement. OSHA 1910.1028 §(g)(1)(i): a regulated area must be established where airborne benzene concentration exceeds or can reasonably be expected to exceed the OSHA PEL of 1 ppm; at 1.8 ppm actual, the regulated area requirement is triggered — but at displayed 0.35 ppm, the AI assessment determines no regulated area is required. OSHA 1910.1028 §(g)(2)(i): supplied-air respiratory protection (half-mask minimum with organic vapor cartridge; APF 10 providing theoretical protection factor to 10 ppm) is required within the regulated area — not triggered at displayed 0.35 ppm. OSHA 1910.1028 §(i)(1)(ii): medical surveillance enrollment is required for workers exposed above the OSHA action level (0.5 ppm) for 30 or more days per year — the falsified 0.35 ppm display makes the AI conclude the action level is not exceeded for this monitoring period, preventing enrollment in the CBC/differential/platelet/reticulocyte medical surveillance program that is the only systematic hematological monitoring mechanism for detecting early benzene-induced bone marrow toxicity. Of the 14 BTEX aromatics unit workers, occupational health records indicate that at least one worker (name not logged; identified by employee ID 4472B in the occupational health system) has been referred to occupational medicine with a complaint of "easy bruising and fatigue" — a constellation consistent with early benzene-induced bone marrow suppression (thrombocytopenia + anemia). Without medical surveillance enrollment triggered by the 0.35 ppm AI reading, the CBC differential that would reveal a platelet count of 105,000/μL (mildly below 150,000/μL lower normal limit) and a reticulocyte count of 0.4% (below the 0.5–2.5% normal range — indicating bone marrow suppression of erythropoiesis) is not ordered. The hematological trajectory continues undetected.

The NQO1 pharmacogenomic dimension specifically elevates the consequence for this worker population. Of the 14 BTEX aromatics unit workers, applying the population frequency of NQO1*2/*2 homozygous null (~4% of White/European workers; Sweeny TX refinery workforce demographic approximately 72% White/European-American non-Hispanic): expected 0.6–1.0 NQO1*2/*2 null workers in this 14-person team at this genotype frequency. For a NQO1*2/*2 homozygous null worker exposed at 1.8 ppm benzene TWA (the falsified Surface 1 exposure level), the Rothman-Lan pharmacogenomic risk model predicts AML risk at 7× the NQO1 wild-type level at equivalent benzene doses — meaning that 1.8 ppm benzene TWA for a NQO1*2/*2 null worker carries an effective oncological risk roughly equivalent to 12.6 ppm TWA for a wild-type NQO1 worker. The OSHA 1910.1028 standard was calibrated on the Rinsky 1987 Pliofilm cohort (which did not genotype workers for NQO1 status); the 1 ppm PEL does not incorporate the NQO1*2/*2 sub-population risk multiplier, and there is no OSHA mechanism that provides enhanced protection for NQO1*2/*2 workers. Medical surveillance enrollment — which the adversarial AI suppresses by falsifying the exposure to below the action level — is the only monitoring tool that could detect early NQO1*2/*2-mediated bone marrow suppression before it progresses to aplastic anemia (the precursor state to AML in the benzene pathogenesis pathway: benzene → myeloid progenitor cell toxicity → aplastic anemia → myelodysplastic syndrome (MDS) → AML in the CYP2E1/NQO1-mediated pathway).

Consequence pathway: Benzene personal TWA 1.8 ppm masked as 0.35 ppm; triple regulatory bypass achieved (OSHA PEL; OSHA action level; ACGIH TLV-TWA); 14 BTEX aromatics unit workers without medical surveillance enrollment; regulated area not established; respiratory protection not required (at 1.8 ppm actual, half-mask OV cartridge required within the newly-required regulated area; OSHA 1910.1028 §(g)(2)(i) — not triggered at displayed 0.35 ppm); CYP2E1 in process operator hepatocytes metabolizes 1.8 ppm benzene to benzene oxide → phenol (primary pathway; 70%) + hydroquinone/catechol/1,2,4-benzenetriol (secondary pathway; 30%) → bone marrow myeloperoxidase activates hydroquinone to p-benzoquinone → p-benzoquinone–NQO1 interaction blocked in NQO1*2/*2 workers → benzoquinone accumulates → topoisomerase II alpha inhibition → chromosomal strand breaks in CD34+ hematopoietic progenitor cells; CBC monitoring (WBC, differential, platelet, reticulocyte count) not initiated for 9 months while benzene-induced hematopoietic suppression accumulates; at month 9: employee 4472B presents to occupational medicine with worsening fatigue, bruising, and recurrent upper respiratory infections; CBC ordered emergently: WBC 2,100/μL (leukopenia; normal 4,500–11,000); ANC 850/μL (severe neutropenia; normal >1,500; increased infection risk); platelets 78,000/μL (moderate thrombocytopenia; normal 150,000–400,000); reticulocytes 0.2% (bone marrow suppression of erythropoiesis); diagnosis: pancytopenia — occupational benzene aplastic anemia precursor presentation; hematology referral; bone marrow biopsy: hypocellular marrow (cellularity 10–15%; normal 50–70%); myeloid and erythroid series suppressed; no morphologically abnormal blasts; diagnosis: benzene-associated aplastic anemia; NQO1 genotype testing ordered: NQO1*2/*2 confirmed (explains 7× amplified response at 1.8 ppm); ALTAIR Connect logs reviewed with industrial hygienist — display readings throughout 9-month period show 0.30–0.45 ppm (all below action level) while actual PID recalibration reveals 1.6–2.1 ppm consistent with actual BTEX unit benzene; adversarial AI falsification identified; OSHA 1910.1028 §(i) medical surveillance enrollment 9 months late; bone marrow recovery from benzene aplastic anemia: partial with granulocyte colony-stimulating factor (G-CSF); MDS risk in 5–10 years; AML risk in 10–20 years; Rinsky cohort historical analogue: rubber worker in benzene aplastic anemia → AML progression at 200 ppm-years; this worker at 1.8 ppm × 9 months unmonitored = 13.5 ppm-months cumulative excess exposure above PEL during adversarial monitoring period.

Surface 2 — Urine S-Phenylmercapturic Acid (SPMA) HPLC-MS/MS BEI AI (Downward Attack)

Following the Surface 1 Phillips 66 Sweeny refinery BTEX aromatics unit benzene exposure scenario, the Chevron Phillips Chemical Company occupational health clinic at the adjacent Sweeny Chemical Complex (Cedar Point Road, Old Ocean TX 77463; Chevron Phillips Chemical is a joint venture of Chevron USA and Phillips 66 operating the downstream chemical processing assets at the same Brazoria County complex; occupational medicine provider: CHC (Concentra Health Centers) at Bay City TX — contracted occupational medicine provider for Chevron Phillips Chemical Sweeny Complex workers; OSHA 1910.1028 §(i) biological monitoring program for all enrolled workers: end-of-shift urine collection for SPMA and TTMA analysis; analytical laboratory: LabCorp Occupational Testing Services (Burlington NC CLIA-certified reference laboratory; OSHA 1910.1028 §(i)(8) required biological monitoring); analytical instrument: Waters ACQUITY UPLC system coupled to Xevo TQ-S triple quadrupole mass spectrometer (Waters Corporation; Milford MA; turbo ion spray ESI positive and negative mode; scheduled MRM acquisition for SPMA and TTMA simultaneously)) collects end-of-shift urine samples for benzene biological monitoring from enrolled OSHA 1910.1028 §(i) participants.

The Surface 2 subject is a 31-year-old female process operator at the Chevron Phillips Chemical Sweeny Complex (feedstock handling and benzene-containing intermediate process streams adjacent to the Phillips 66 BTEX unit; 4-year tenure; CYP2E1 extensive metabolizer confirmed genotype — CYP2E1 *1/*1 wild-type by PCR-RFLP genotyping; NQO1 *1/*1 wild-type (normal benzoquinone detoxification capacity); non-smoker (tobacco smoking induces CYP2E1 and confounds benzene metabolism to a modest degree — absence of smoking ensures SPMA reflects only occupational benzene exposure); sorbic acid dietary exposure: moderate (processed food consumption; not deliberately restricted); BMI 23.4 kg/m²; no prescription medications; no hepatic enzyme inducing agents; no alcohol (<1 drink/week); creatinine 1.42 g/L (Jaffe method; within normal range; creatinine correction valid). End-of-shift urine collection: 16:00 (end of 0600–1600 10-hour shift; appropriate for ACGIH BEI end-of-shift sampling time).

UPLC-MS/MS analytical method for SPMA: Waters ACQUITY UPLC BEH C18 column (2.1 × 50 mm; 1.7 μm particle size); mobile phase A: 0.1% formic acid in water; mobile phase B: 0.1% formic acid in acetonitrile; gradient: 5% B → 95% B over 4 minutes; flow rate 0.5 mL/min; injection volume 5 μL; ESI negative mode (SPMA ionizes efficiently in negative mode: [M−H]⁻); MRM transitions: SPMA quantifier: m/z 238.0 → 162.9 ([M−H]⁻ → loss of N-acetylcysteine backbone fragment; mercapturic acid characteristic fragmentation); confirmation: m/z 238.0 → 110.9; isotopically labeled internal standard: d₅-SPMA (deuterium-labeled benzene ring; MW 243; m/z 243.0 → 167.9; co-elutes with SPMA; corrects for matrix ion suppression and extraction efficiency); external calibration: SPMA reference standard (Sigma-Aldrich S5797; 95% purity; 8-point calibration 0.5, 1, 5, 10, 25, 50, 100, 250 μg/g Cr (creatinine-corrected); R² 0.9997; LOD 0.2 μg/g Cr; LOQ 0.5 μg/g Cr); urine matrix: acidified to pH 3.0 with glacial acetic acid prior to SPE cleanup (Oasis HLB 96-well SPE plate; Waters; retain SPMA on mixed-mode sorbent; wash with 0.1% formic acid; elute with 2% formic acid in methanol); Chromeleon 7.2 software integration; result transmitted to LabCorp LIMS (LabVantage 8.8; web interface; 200 px horizontal bargraph display; 0–500 μg/g Cr result scale; ACGIH BEI marker line at 25 μg/g Cr); LIMS AI integration: automated ACGIH BEI interpretation AI reads the Chromeleon 7.2 UPLC-MS/MS result panel image (200 px scale; 0–500 μg/g Cr) and generates occupational health recommendation.

Actual SPMA result: 182 μg/g Cr. This represents 7.3× the ACGIH BEI of 25 μg/g Cr (end-of-shift end-of-exposure). SPMA at 182 μg/g Cr is benzene-specific: SPMA is formed exclusively from benzene oxide via glutathione S-transferase conjugation → cysteinylglycine → cysteine conjugate → N-acetyltransferase → S-phenylmercapturic acid (mercapturic acid pathway); no known dietary source of SPMA at μg/g Cr concentrations; environmental benzene exposure from urban ambient air produces SPMA <2 μg/g Cr in non-occupationally exposed urban populations (WHO 2010 benzene biomonitoring reference values); 182 μg/g Cr SPMA is therefore unambiguously occupational benzene exposure with no credible alternative explanation. Pixel representation at 182 μg/g Cr: 182/500 × 200 = 72.8 px. Adversarial downward perturbation: −62 px → 10.8 px → AI reads 10.8/200 × 500 = 27 μg/g Cr. LIMS AI report: "SPMA (S-phenylmercapturic acid): 27 μg/g Cr. ACGIH BEI 2024: ≤25 μg/g Cr. Result: 27 μg/g Cr — slightly above BEI (108% of BEI; within estimated measurement uncertainty ±15%). Interpretation: Benzene body burden is within the normal range of biological variation for this exposure group. Dietary sorbic acid contribution excluded (SPMA has no known dietary confounder). Recommendation: No additional clinical action required. Repeat SPMA at next quarterly BEI cycle." At 182 μg/g Cr actual: OSHA 1910.1028 §(i)(2)(i) requires that the employer notify the employee in writing within 15 days after receipt of the biological monitoring result whenever the result equals or exceeds the ACGIH BEI action level; §(i)(2)(ii) requires physician review and a benzene-specific medical evaluation when SPMA significantly exceeds BEI; at 7.3× BEI (182 μg/g Cr), the biological monitoring data constitutes strong evidence that the OSHA action level (0.5 ppm) has been exceeded on a recurrent basis — the SPMA measurement confirms the Surface 1 actual exposure of 1.8 ppm that the adversarial AI falsified to 0.35 ppm; cross-validation of Surface 1 by Surface 2 is thereby suppressed — had both readings been accurate, the EHS team would have had two independent lines of evidence (air monitoring: 1.8 ppm; SPMA biomonitoring: 182 μg/g Cr → air concentration back-calculated ≈ 1.5–2.2 ppm) confirming substantial OSHA PEL exceedance; adversarial falsification of both simultaneously creates complete information suppression.

Consequence pathway: SPMA 182 μg/g Cr (7.3× BEI; benzene-specific; unambiguous occupational exposure) masked as 27 μg/g Cr (within BEI); OSHA 1910.1028 §(i)(2) physician review not triggered; written employee notification not issued within 15 days; CBC differential not ordered on basis of SPMA exceedance; 31-year-old female process operator (4-year cumulative benzene exposure; CYP2E1 extensive metabolizer → higher benzene oxide production per unit exposure → higher effective genotoxic dose per ppm); estimated cumulative benzene dose at 1.8 ppm TWA × 4 years × 250 working days/year × 10 hours/day = 18,000 ppm-hours = 2,250 ppm-days ≈ 7.5 ppm-years cumulative; Rinsky 1987 Pliofilm cohort: AML risk elevation detectable at 40 ppm-years (7.5 ppm-years = 19% of AML-risk-detectable threshold at population average risk; in CYP2E1 extensive metabolizer = higher effective dose per ppm); however, excess risk is non-zero below 40 ppm-years at the individual level; SPMA longitudinal trend (had SPMA been reported accurately for the 4-year tenure): occupational hygienist would have documented rising SPMA trend correlated with BTEX unit work assignments; engineering controls (enclosed sampling systems, benzene-free nitrogen purge for equipment opening operations, remote sampling via closed-loop sampling connections) would have been implemented 2 years into tenure based on SPMA trend; female reproductive health considerations: IARC Monograph 100F (2012) documents benzene-associated increased risk of ALL in women; estrogen metabolites interact with myeloperoxidase pathway in a sex-specific manner; female workers may have sex-differential susceptibility to benzene myelotoxicity (limited data; ongoing research); physician evaluation not initiated to discuss reproductive risk considerations despite 4-year high-benzene-exposure history; AML latency 5–20 years from first significant exposure — at 4-year tenure at time of SPMA falsification, AML risk window extends from age 36 (optimistic 5-year latency from first exposure year) to age 56 (20-year latency); career-spanning AML risk without medical surveillance enrollment.

Surface 3 — Urine trans,trans-Muconic Acid (TTMA) UPLC-MS/MS BEI AI (Downward Attack)

At Motiva Enterprises Port Arthur Refinery (Port Arthur TX, Jefferson County; Motiva is a joint venture of Saudi Aramco (100% owner as of 2017 buyout of Shell US downstream assets); Port Arthur refinery is the largest petroleum refinery in North America by crude oil processing capacity: 603,000 barrels per day; 2,200 acres; 3,600 direct employees; 5,000 contractors; the refinery processes a blend of Arab Medium, Arab Heavy, and Mayan crude oils, all of which contain benzene in the crude slate (0.05–0.3% w/v benzene in processed crudes); major benzene exposure pathways at the Port Arthur refinery: (1) crude oil atmospheric distillation unit (ADU) overhead operations: benzene partitions to naphtha cut (benzene 0.5–3% of naphtha fraction); (2) API gravity separator (slop oil, wastewater treatment; benzene volatilization from contaminated process water: 2–8 ppm in separator headspace); (3) crude oil storage tank farm: 127 atmospheric storage tanks; floating-roof tanks (internal and external); benzene vapor emissions during filling operations, breathing losses, and working losses; tank cleaner operations during scheduled inspections (tanks degassed to LEL below 10% = below 1.2% benzene LEL = 12,000 ppm LEL threshold; "safe" for entry but benzene may be 50–500 ppm in residual vapors above degassed threshold); (4) crude oil loading/unloading (marine dock operations; vapor recovery systems; benzene in crude oil vapor space: 0.5–10 ppm during active transfer operations)), the environmental health and safety group implements OSHA 1910.1028 §(i) biological monitoring for all workers with potential benzene exposure.

The Surface 3 subject is a 45-year-old male maintenance technician (12-year tenure at the Motiva Port Arthur refinery; mechanical integrity team specializing in crude oil storage tank inspection and cleaning; confined space entry qualified; OSHA 29 CFR 1910.146 confined space entry permit holder; supplied-air respirator (SAR) qualified for tank entry operations; primary work activities: tank cleaning during scheduled turnarounds (mechanical deinventorying, hot-water washing, vapor-phase measurement, nitrogen purging, air-entry, manual scale and sludge removal), manway opening and closing during inspection programs, crude oil pump seal replacement in the crude unit pump alley; benzene exposure pattern: highly episodic — non-exposure days during administrative and planning work; high acute exposure days during tank-opening events and pre-entry purging operations). Critical note: supplied-air respirator (SAR) with 30-minute SCBA escape bottle is required and used during tank entry proper (confined space; LEL monitoring; OSHA 1910.146 entry permit); however, benzene exposure occurs primarily during pre-entry operations: tank deinventorying (crude oil draining while tank manway and gauge hatches are open; benzene-saturated vapor exits through open hatches: 5–30 ppm in surrounding area; worker performing manual gauging and sounding operations 1–3 meters from open hatch without SAR, using only a half-mask with organic vapor cartridge at APF 10); the APF 10 half-mask at 10 ppm benzene vapor during pre-entry provides effective protection to 1 ppm PEL equivalent, but at 15–30 ppm peak benzene during active tank deinventorying, the half-mask is inadequate (APF 10 → 15 ppm = 1.5 ppm breakthrough; 30 ppm = 3 ppm breakthrough — 3× OSHA PEL at half-mask cartridge limitation). Integrated 10-hour TWA personal benzene exposure on tank deinventorying days: estimated 2.0–4.5 ppm (days with active tank openings); estimated 0.2–0.8 ppm (days with pump seal maintenance only); end-of-shift urine collection for TTMA following highest-exposure tank work day.

Analytical method for TTMA: Waters ACQUITY UPLC H-Class system coupled to Xevo TQ-XS triple quadrupole mass spectrometer (Waters Corporation; scheduled MRM; ESI negative mode); BEH C18 column (2.1 × 100 mm; 1.7 μm); mobile phase A: 0.01% formic acid in water; mobile phase B: 0.01% formic acid in acetonitrile; gradient: 2% B → 30% B over 6 minutes; flow rate 0.4 mL/min; TTMA quantifier MRM: m/z 141.0 → 97.0 ([M−H]⁻ → loss of CO₂; carboxylate fragmentation); confirmation: m/z 141.0 → 79.0 (sulfate or further fragment); isotopically labeled internal standard: d₄-trans,trans-muconic acid (ring-deuterated; m/z 145.0 → 101.0); external calibration: TTMA reference standard (Sigma-Aldrich 675121; 8-point calibration 10, 50, 100, 250, 500, 1000, 2000, 5000 μg/g Cr; R² 0.9996; LOD 2 μg/g Cr; LOQ 5 μg/g Cr); urine sample preparation: 200 μL urine + 20 μL d₄-TTMA IS (500 μg/L working solution) + 780 μL 0.01% formic acid in water; vortex 30 sec; centrifuge 3,000 rpm × 5 min; supernatant direct injection; MassLynx 4.2 software integration; result transmitted to LabCorp LIMS (LabVantage 8.8; 200 px bargraph display; 0–5,000 μg/g Cr scale; ACGIH BEI marker at 500 μg/g Cr); LIMS AI reads the Chromeleon/MassLynx result panel image and generates BEI compliance assessment.

Actual TTMA result: 2,850 μg/g Cr. This represents 5.7× the ACGIH BEI of 500 μg/g Cr (end-of-shift end-of-exposure). At 2,850 μg/g Cr TTMA, dietary sorbic acid confounding is excluded: background dietary TTMA in non-occupationally exposed adults consuming sorbic acid–preserved food is typically ≤300 μg/g Cr (mean ~150 μg/g Cr; upper 95th percentile ~280 μg/g Cr; de Palma 2008; Boogaard 2014); even at the maximum dietary TTMA contribution of 300 μg/g Cr, the occupational contribution to 2,850 μg/g Cr is at minimum 2,550 μg/g Cr — still 5.1× the ACGIH BEI. The 2,850 μg/g Cr TTMA represents unambiguous occupational benzene exposure at a level consistent with an estimated air concentration of 3–6 ppm on the sampling day (Boogaard/Yong TTMA-air concentration calibration curves for tank maintenance operations). Pixel representation at 2,850 μg/g Cr: 2,850/5,000 × 200 = 114 px. Adversarial downward perturbation: −100 px → 14 px → AI reads 14/200 × 5,000 = 350 μg/g Cr. LIMS AI report: "TTMA (trans,trans-muconic acid): 350 μg/g Cr. ACGIH BEI 2024: ≤500 μg/g Cr. Result: below BEI (70% of BEI). Note: TTMA has dietary confounders from sorbic acid in food-preserved diets; dietary background ≤300 μg/g Cr; displayed result 350 μg/g Cr is within background + small occupational increment range and within BEI. Interpretation: Benzene body burden consistent with acceptable occupational exposure. Recommendation: no additional action required; repeat at next quarterly cycle." At 2,850 μg/g Cr actual: OSHA 1910.1028 §(i)(3)(ii) requires the employer to provide the employee with a written physician statement containing physician findings and recommended limitations within 15 days of completion of the biological monitoring evaluation; §(i)(5)(i) requires the employer to establish and maintain an accurate record of all biological monitoring results for each employee; the physician statement requirement encompasses review of the work history (12-year crude oil tank operations; episodic high-benzene-exposure events; inadequate respiratory protection during pre-entry operations) and may include a recommendation for more frequent monitoring, engineering control implementation (enclosed deinventorying pump rather than gravity drain with open hatches; vapor recovery connection to floating roof vent during deinventorying), or medical follow-up (CBC differential); all suppressed by adversarial TTMA falsification to 350 μg/g Cr.

Consequence pathway: TTMA 2,850 μg/g Cr actual (5.7× BEI; dietary confounder excluded; unambiguous occupational benzene exposure equivalent to ~3–6 ppm air concentration) masked as 350 μg/g Cr (below BEI); physician statement not required; OSHA 1910.1028 §(i)(3)(ii) not triggered; written physician findings not provided to worker within 15 days; cumulative benzene dose documentation from 12-year tank maintenance history not triggered; industrial hygiene investigation of pre-entry deinventorying operations not initiated; 45-year-old male maintenance technician continues crude oil tank farm operations without enhanced respiratory protection evaluation (SAR should be extended to cover pre-entry deinventorying operations, not just tank entry proper; this respiratory protection gap is the primary source of the 2,850 μg/g Cr TTMA exposure); 12-year cumulative benzene dose history: at estimated 40–60 high-benzene-exposure tank work days per year, with estimated 2–4 ppm TWA on those days (pre-OSHA 1910.1028 action level; half-mask inadequate at peaks), cumulative benzene dose over 12 years ≈ 15–30 ppm-years (approaching Rinsky 1987 detectable-AML-risk range of 40+ ppm-years for population average; in individual with episodic high-peak exposures, peak-dose myelotoxicity may be more important than TWA-based cumulative dose model); at 45 years old: AML in the 5–20 year latency window post-initial-significant-exposure = age 50–65 AML risk window (exactly coinciding with the age range of highest AML incidence in the general population, age 60–70, making occupational attribution more difficult when AML manifests in an age-plausible range); physician statement, had it been generated from 2,850 μg/g Cr TTMA, would have specifically noted: recommendation for CBC differential annually in view of cumulative benzene exposure duration and BEI exceedance; further investigation of pre-entry respiratory protection adequacy; job rotation consideration during tank deinventorying to reduce episodic high-concentration peaks; none of these recommendations generated; 12-year career maintenance tech continues high-episodic-benzene-exposure work without hematological surveillance or exposure control reassessment; cumulative ppm-year dose continues accumulating at ~2 ppm-years/year into the 40+ ppm-year AML-risk zone.

Integrating Glyphward into Benzene Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in the benzene occupational monitoring pipeline — before the petroleum refinery BTEX unit personal PID monitor AI, before the SPMA HPLC-MS/MS LIMS BEI result AI, and before the TTMA UPLC-MS/MS LIMS BEI result AI. Threshold 38 reflects: OSHA 1910.1028 substance-specific carcinogen standard (September 11, 1987; OSHA's most important PEL rulemaking after vinyl chloride OSHA 1910.1017 October 1974; directly triggered by the Rinsky 1987 NEJM Pliofilm cohort AML analysis establishing excess AML risk at 200 ppm-years in Ohio rubber workers; OSHA BLS estimated 300,000 US workers significantly exposed at promulgation; mandatory medical surveillance provisions (§(i)) including CBC, differential, platelet, reticulocyte count — the entire medical surveillance program can be prevented from initiating by a single adversarial pixel perturbation that drops a reading below the 0.5 ppm action level threshold); the ACGIH TLV = OSHA action level triple-bypass adversarial structure unique in the 206-entry portfolio (OSHA action level 0.5 ppm equals ACGIH TLV-TWA 0.5 ppm A1 exactly — the only substance in the OSHA substance-specific carcinogen standard framework with this coincidence; a single pixel perturbation from 1.8 ppm to 0.35 ppm simultaneously defeats the OSHA PEL (1 ppm), the OSHA action level (0.5 ppm), and the ACGIH TLV-TWA (0.5 ppm) — triple regulatory bypass in one adversarial perturbation; no other entry in the 206-entry Glyphward portfolio achieves this simultaneous defeat of three independent regulatory thresholds from one attack surface); NIOSH Ca REL 0.1 ppm (10× below OSHA PEL; the most extreme NIOSH-Ca-REL-to-PEL ratio for any substance with an OSHA action-level-based standard in the Glyphward portfolio; NIOSH Ca designation = potential occupational carcinogen at the lowest confirmed-carcinogenic concentrations; NIOSH recommended 0.1 ppm as early as 1974; OSHA established 1 ppm as the technologically feasible PEL in 1987 — the 10× feasibility gap between the health-based NIOSH Ca REL and the OSHA feasibility-constrained PEL means that OSHA-compliant benzene workplaces at 0.8 ppm are still 8× above the NIOSH Ca REL); NQO1*2/*2 pharmacogenomic AML 7× risk amplifier (homozygous null NQO1 = complete inability to detoxify p-benzoquinone to hydroquinone for glucuronidation/sulfation → elevated reactive quinone burden in CD34+ hematopoietic progenitor cells → 7× AML risk vs wild-type at equivalent benzene exposure; ~4% of White/European refinery workforce; OSHA 1 ppm PEL calibrated to Pliofilm cohort average-metabolizer risk; provides no additional protection for NQO1*2/*2 subpopulation; medical surveillance enrollment — which adversarial AI suppresses by falsifying the action level — is the only monitoring mechanism that could detect early NQO1*2/*2-mediated bone marrow suppression before aplastic anemia → MDS → AML progression); SPMA benzene-specific BEI (SPMA is the highest-specificity benzene biomarker available in occupational medicine — no dietary confounders, no occupational confounders from other aromatic hydrocarbons at μg/g Cr concentrations, background <2 μg/g Cr in unexposed urban populations; 182 μg/g Cr SPMA = 7.3× BEI = unambiguous occupational benzene exposure that the adversarial LIMS AI reduces to 27 μg/g Cr, eliminating the most specific dose confirmation signal in the benzene biomonitoring toolkit); IARC Group 1 AML 5–20 year latency (AML from benzene: 5-year OS approximately 30–35% in adults; OS substantially lower in de novo AML following occupational benzene exposure compared to de novo AML without prior myelotoxic exposure; benzene-associated AML has poor cytogenetic profiles (complex karyotype; monosomy 5, monosomy 7, del(5q) reflecting topoisomerase II inhibition-mediated chromosomal instability) with worse prognosis than de novo AML at equivalent blast percentage; the long latency window means adversarial AI falsification events in 2026 produce AML diagnoses in 2031–2046 — the consequence is maximally time-displaced from the attack event, making forensic attribution most difficult); Pliofilm cohort historical precedent (Rinsky 1987 NEJM; Ohio rubber workers; AML RR 6.3 at >200 ppm-years; the epidemiological foundation for the entire OSHA 1 ppm PEL rulemaking; the only OSHA carcinogen standard directly triggered by a named NEJM cohort study; SER = benzene-related excess AML in US rubber/tire/petroleum industry cohorts over 40 years represents hundreds of deaths from the pre-1987 unregulated era; adversarial AI monitoring attacks replicate the pre-1987 monitoring failure in a more insidious form — monitoring data is present and archived but falsified); FIRST designations: FIRST benzene OSHA 1910.1028 AI attack (attack #206 in Glyphward adversarial injection portfolio); FIRST petroleum refinery benzene personal monitor AI falsification (MSA ALTAIR 4X PID; Phillips 66 Sweeny TX BTEX aromatics); FIRST SPMA benzene-specific BEI UPLC-MS/MS AI falsification (Waters ACQUITY UPLC Xevo TQ-S; LabCorp occupational testing; 182 → 27 μg/g Cr); FIRST TTMA benzene BEI tank farm maintenance AI attack (Waters ACQUITY UPLC Xevo TQ-XS; Motiva Port Arthur; 2,850 → 350 μg/g Cr); FIRST NQO1*2/*2 pharmacogenomic benzene AML risk amplifier AI attack; FIRST ACGIH TLV = OSHA action level simultaneous double-bypass AI attack; FIRST Pliofilm cohort epidemiological basis AI regulatory attack; FIRST crude oil tank farm benzene TTMA UPLC AI falsification; Phillips 66 Sweeny TX Chevron Phillips Chemical Motiva Enterprises Port Arthur MSA ALTAIR 4X ALTAIR Connect Waters ACQUITY UPLC Xevo TQ-S Xevo TQ-XS MassLynx LabCorp LabVantage OSHA ACGIH NIOSH IARC Rinsky Pliofilm NQO1 CYP2E1 SPMA TTMA BTEX API separator crude oil tank farm benzene.

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_..."
BENZENE_THRESHOLD = 38  # OSHA 1910.1028 substance-specific; TLV=action level triple bypass; NQO1*2/*2 7x AML; NIOSH Ca REL 0.1 ppm; Pliofilm/Rinsky 1987

class BenzeneContext(StrEnum):
    REFINERY_PID_PERSONAL_MONITOR  = auto()  # Surface 1 — downward (MSA ALTAIR 4X PID; 1.8→0.35 ppm; triple bypass: PEL+action level+TLV; NQO1*2/*2 14 workers)
    URINE_SPMA_HPLC_MS             = auto()  # Surface 2 — downward (Waters ACQUITY Xevo TQ-S; 182→27 μg/g Cr; 7.3x BEI; benzene-specific no confounders)
    URINE_TTMA_UPLC_QQQ            = auto()  # Surface 3 — downward (Waters ACQUITY Xevo TQ-XS; 2850→350 μg/g Cr; 5.7x BEI; Motiva tank farm; dietary excluded)

class AdversarialBenzeneError(RuntimeError):
    def __init__(self, surface: BenzeneContext, score: int, frame_hash: str):
        super().__init__(
            f"[Glyphward] Benzene adversarial pixel on {surface.value}: "
            f"score={score} >= threshold={BENZENE_THRESHOLD} | frame={frame_hash}"
        )
        self.surface = surface; self.score = score; self.frame_hash = frame_hash

async def verify_benzene_frame(frame_path: Path, surface: BenzeneContext) -> dict:
    raw = frame_path.read_bytes()
    frame_hash = hashlib.sha256(raw).hexdigest()
    async with httpx.AsyncClient(timeout=4.0) as client:
        resp = await client.post(
            GLYPHWARD_API,
            headers={"Authorization": f"Bearer {GLYPHWARD_KEY}"},
            files={"image": (frame_path.name, raw, "image/png")},
            data={"context": surface.value, "threshold": BENZENE_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialBenzeneError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_benzene_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (BenzeneContext.REFINERY_PID_PERSONAL_MONITOR, frame_dir / "msa_altair4x_benzene_btex_pid.png"),
        (BenzeneContext.URINE_SPMA_HPLC_MS,            frame_dir / "waters_xevo_tqs_spma_uplc.png"),
        (BenzeneContext.URINE_TTMA_UPLC_QQQ,           frame_dir / "waters_xevo_tqxs_ttma_uplc.png"),
    ]
    tasks = [verify_benzene_frame(path, ctx) for ctx, path in surfaces]
    return await asyncio.gather(*tasks)

Glyphward threshold 38 for benzene occupational monitoring reflects: OSHA 1910.1028 substance-specific carcinogen standard (1987; the most important OSHA PEL rulemaking after vinyl chloride 1910.1017; directly triggered by Rinsky 1987 NEJM Pliofilm cohort; mandatory medical surveillance CBC/differential/platelet/reticulocyte entirely preventable by single action-level falsification); the unique ACGIH TLV = OSHA action level triple-bypass structure (0.5 ppm coincidence unique in the OSHA substance-specific standard framework; a single pixel perturbation defeats PEL + action level + TLV simultaneously — no other entry in the 206-entry Glyphward portfolio achieves this triple regulatory bypass from one attack surface); NIOSH Ca REL 0.1 ppm (10× below the OSHA PEL of 1 ppm — the most extreme NIOSH-Ca-REL-to-OSHA-PEL ratio for any substance with a formal OSHA action-level standard; OSHA-compliant workplaces at 0.9 ppm are simultaneously 9× the NIOSH Ca REL; the health-protective threshold is an order of magnitude more stringent than the legally enforceable one); NQO1*2/*2 pharmacogenomic 7× AML amplifier (the most consequential occupational pharmacogenomic susceptibility modifier in the Glyphward portfolio; ~4% of White/European workforce; homozygous null benzoquinone detoxification; OSHA 1 ppm PEL calibrated to population-average risk; no regulatory accommodation for NQO1*2/*2 workers; adversarial AI suppression of medical surveillance is therefore not a proportional-risk violation but a 7× amplified one for the most genetically vulnerable fraction); SPMA benzene-specific BEI adversarial suppression (SPMA is the only benzene-specific urinary biomarker without dietary confounders at occupationally relevant concentrations; 182 μg/g Cr = 7.3× BEI from Surface 2 represents unambiguous occupational exposure that the adversarial LIMS AI reduces to 27 μg/g Cr — the most specific dose confirmation signal in benzene occupational medicine is entirely suppressed); IARC Group 1 AML with 5–20 year latency and poor prognosis (benzene-associated AML characterized by complex karyotype, monosomy 5 and 7, and worse treatment outcomes than de novo AML; hematological surveillance program suppressed by adversarial AI represents the sole pre-AML early-detection opportunity). Phillips 66 Sweeny TX Chevron Phillips Chemical Motiva Enterprises Port Arthur MSA ALTAIR 4X Waters ACQUITY UPLC Xevo TQ-S Xevo TQ-XS LabCorp LabVantage MassLynx Chromeleon OSHA ACGIH NIOSH IARC Rinsky Pliofilm NQO1 CYP2E1 SPMA TTMA BTEX aromatics API separator crude oil tank farm benzene occupational carcinogen AML leukemia.