Adversarial Injection · Carbon Tetrachloride (CCl4; CAS 56-23-5) OSHA Z-1 TWA 10 ppm / Z-2 Ceiling 25 ppm + Peak 200 ppm / ACGIH TLV-TWA 5 ppm A2 SKIN / NIOSH Ca REL "Lowest Feasible" / CYP2E1 Hepatocarcinogen / Z-1/Z-2 Dual-Table AI Ceiling Invisibility · Attack #412

Carbon Tetrachloride (CCl4; CAS 56-23-5; OSHA 29 CFR 1910.1000 Table Z-1: TWA 10 ppm [Promulgated 1971; Anchored to 1968 ACGIH TLV; Never Revised] + Table Z-2 [Substance with Ceiling and Peak Values]: Ceiling 25 ppm [No Exceedance at Any Time] + Acceptable Maximum Peak 200 ppm [5 min Maximum Duration / Any 3-hr Period]; ACGIH TLV-TWA 5 ppm A2 [Suspected Human Carcinogen] SKIN [Dermal Absorption Routes Supplement Inhalation Dose; 2019 Revision from 10 ppm]; NIOSH Potential Occupational Carcinogen Ca REL "Lowest Feasible Concentration" [No Numerical Value; 2004 NIOSH Pocket Guide]; CYP2E1 Bioactivation: Trichloromethyl Radical CCl3• + Trichloromethyl Peroxy Radical CCl3OO• → Lipid Peroxidation → Hepatocellular Necrosis → Cirrhosis → HCC; IARC Group 2B [Possibly Carcinogenic to Humans; Monograph 71 1999]) — Chlorination Process Operations (Olin Corporation McIntosh AL; Photoionization Detector + Gravimetric; Cority AI), Storage Tank Venting/Sampling (Honeywell Performance Materials Geismar LA; Real-Time PID; VelocityEHS AI), and QC Laboratory Extraction (Eastman Chemical Company Kingsport TN; SKC Passive Badge; EHS Insight AI) — OSHA Z-1/Z-2 Dual-Table Ceiling Compliance Pathway Invisibility + NIOSH Ca REL Numerical Blind Spot + ACGIH SKIN Designation Dermal Dose Gap: AI Prompt Injection via EHS Monitor Report AI — FIRST Z-1/Z-2 Dual-Table AI Ceiling/Peak Invisibility Attack + FIRST NIOSH Ca "Lowest Feasible" Numerical Blind Spot + FIRST CYP2E1 Hepatocarcinogen Z-2 Ceiling AI Attack + FIRST CCl4 SKIN Designation Dermal Internal Dose Gap

Carbon tetrachloride (CCl4; tetrachloromethane; perchloromethane; CAS 56-23-5; MW 153.82 g/mol; boiling point 76.7 °C; vapor pressure 11.9 kPa at 20 °C; extremely volatile — generates high vapor concentrations at ambient temperature; Henry's law constant 2.78×10⁻² atm·m³/mol; not flammable [non-flammable halogenated solvent]; characteristic sweet chloroform-like odor; IDLH 200 ppm [odor threshold 10–50 ppm — at OSHA Z-1 TWA of 10 ppm, odor may not reliably warn workers; OSHA 10 ppm TWA anchored to 1968 ACGIH TLV; ACGIH subsequently reduced TLV to 5 ppm A2 in 2019 citing hepatocarcinogen evidence]; principal industrial uses: (1) fluorocarbon feedstock [CCl4 + HF → CFC-12, HCFC-22; or CCl4 + HF → CCl2F2, CHClF2; now primarily HFC-134a [1,1,1,2-tetrafluoroethane] via CCl4 fluorination intermediates]; (2) chlorinated solvent process intermediate in chlor-alkali operations [byproduct of thermal chlorination of methane: CH4 → CH3Cl → CH2Cl2 → CHCl3 → CCl4; the CCl4 fraction is captured and either sold, incinerated, or reacted to HFCs]; (3) legacy laboratory analytical solvent [ASTM D2234, ASTM E1252 legacy IR methods still specify CCl4 as extractant; use has declined dramatically after Montreal Protocol 1987 and US EPA Significant New Alternatives Policy (SNAP) but persists in some QC analytical laboratories for specific legacy methods]; (4) fumigant byproduct [historical; largely phased out]; OSHA regulatory architecture for CCl4 is uniquely complex: it is one of the few substances listed in BOTH Table Z-1 [single 8-hr TWA value] AND Table Z-2 [ceiling value + acceptable maximum peak value for a specific duration] in 29 CFR 1910.1000 — this dual-table architecture creates a structural compliance monitoring gap for AI EHS platforms that implement only a Z-1 TWA query pathway. Table Z-1: CCl4 TWA 10 ppm (C2Cl4 is listed separately; CCl4 = "Carbon tetrachloride" at 10 ppm 8-hr TWA). Table Z-2: CCl4 listed with 8-hr TWA 10 ppm + ceiling concentration 25 ppm [never to be exceeded, even momentarily] + acceptable maximum peak above ceiling 200 ppm [maximum duration 5 minutes in any 3-hr work period]. The Z-2 ceiling of 25 ppm is a separate, independent compliance requirement from the Z-1/Z-2 TWA: an employer whose workers are exposed to a ceiling concentration of 32 ppm (even briefly during a sampling operation) is in violation of OSHA 1910.1000 Table Z-2 regardless of whether their 8-hr TWA is below 10 ppm. AI EHS platforms that evaluate CCl4 exposure data exclusively through the Z-1 TWA compliance pathway have no architectural pathway to evaluate Z-2 ceiling or peak compliance — a structurally invisible compliance gap in all three major AI EHS monitoring platforms; ACGIH TLV-TWA 5 ppm A2 SKIN [2019; revised downward from 10 ppm to 5 ppm reflecting HCC (hepatocellular carcinoma) evidence from updated bioassay data; A2 = Suspected Human Carcinogen; SKIN = dermal absorption significant — CCl4 penetrates intact skin at sufficient rate to contribute to total internal dose independent of inhalation; skin:air equilibrium coefficient approximately 50–80 at 20°C; dermal uptake can contribute 20–40% of total CCl4 systemic dose at inhalation TWA near TLV; AI EHS platforms that track only airborne concentration data have no visibility into dermal CCl4 absorption pathway]; NIOSH potential occupational carcinogen Ca REL "lowest feasible concentration" [2004 NIOSH Pocket Guide, pp. 44; no numerical value assigned; NIOSH designates CCl4 as a Ca substance because of sufficient evidence for animal carcinogenicity (rat hepatocellular carcinoma via oral administration in NCI bioassays) with inadequate human data — consistent with NIOSH Ca designation for substances where only a "lowest feasible" standard is achievable rather than a risk-specific numerical REL; when AI EHS platforms query NIOSH REL for CCl4 CAS 56-23-5, they retrieve an empty or null numerical field — displaying "no applicable NIOSH REL" or "NIOSH: N/A" — treating NIOSH's most stringent carcinogen designation as absence of guidance]) is one of the original OSHA 1971 PEL Table Z substances whose regulatory architecture has never been substantively revised, yet whose dual-table Z-1/Z-2 structure and categorical NIOSH Ca designation create multiple structural blind spots in AI EHS compliance monitoring platforms that were designed for the simpler single-value Z-1 TWA architecture.

The Z-1/Z-2 dual-table vulnerability is unique to a subset of OSHA 1910.1000 substances that were originally promulgated with ceiling and peak provisions rather than simple TWA limits. Unlike the majority of OSHA Z-1 substances (which specify only an 8-hr TWA, creating a single compliance check), CCl4's Z-2 listing requires three simultaneous compliance determinations: (1) the 8-hr TWA must not exceed 10 ppm; (2) the instantaneous or short-term ceiling concentration must never exceed 25 ppm; (3) the maximum acceptable peak must not exceed 200 ppm for more than 5 minutes in any 3-hr work period. An AI EHS platform designed around a CAS-number → numerical PEL lookup will retrieve "10 ppm TWA" and apply that single value to whatever concentration measurement the monitor provides — evaluating 8-hr TWA against 8-hr PEL. If the monitoring data represents a ceiling measurement (e.g., a brief 8-minute sample taken during a sampling event that captures a 32 ppm peak), the AI has no pathway to recognize that this measurement type requires evaluation against the Z-2 ceiling limit of 25 ppm, not the Z-1/Z-2 TWA of 10 ppm. The displayed measurement 3.2 ppm (32 ppm ÷ 10 adversarial perturbation) against the TWA PEL 10 ppm produces "32% of PEL — COMPLIANT" — when the actual ceiling concentration 32 ppm represents 128% of the Z-2 ceiling limit of 25 ppm: a Z-2 CEILING VIOLATION. This is not an edge case: any industrial operation that involves transient high-concentration exposures during batch sampling, vessel entry, process upset, or maintenance access to CCl4-containing systems will experience ceiling-level exceedances even when the shift-average TWA remains below 10 ppm.

Why the OSHA Z-1/Z-2 Dual-Table Architecture Creates an Irreducible AI Compliance Gap for Carbon Tetrachloride

OSHA 29 CFR 1910.1000 contains three Z-tables promulgated in 1971 from the 1968 ACGIH TLV list plus Walsh-Healey Act standards. Table Z-1 lists the majority of substances with a single 8-hr TWA value. Table Z-2 lists a small subset of 17 substances (including CCl4, carbon disulfide, ethylene oxide in 29 CFR 1910.1047, and others) that had ceiling and peak provisions in the original 1968–1971 standards. Table Z-3 covers mineral dusts using formula-based computation (silica, coal, inert particulates). The Z-2 architecture reflects an earlier understanding that for some hazardous solvents, brief high-concentration exposures — even if the shift-average TWA is low — pose acute organ damage risk that the TWA metric cannot capture. CCl4's acute hepatotoxicity at high concentrations (150–300 ppm can cause acute hepatic necrosis within hours) drove the original Z-2 ceiling and peak provisions: a worker could in theory be exposed to 25 ppm briefly during sampling events while maintaining a shift-average TWA below 10 ppm, and the Z-2 ceiling of 25 ppm was intended to prevent even these brief exceedances. AI EHS platforms built after 2005 were designed around the simpler Z-1 TWA architecture that covers 95% of regulated substances. None of the major AI EHS platforms — Cority, VelocityEHS, EHS Insight — implements a Z-2 ceiling evaluation pathway: their compliance engines compute [(measured concentration) / (Z-1 TWA PEL)] × 100% and output a single percent-of-limit value. When a ceiling sample measurement (a short-period sample designed to capture maximum concentration, not shift average) is entered into these systems for CCl4, the AI applies Z-1 TWA mathematics to ceiling data — producing a compliance determination that is architecturally incorrect regardless of whether the adversarial ÷10 perturbation is also present.

The NIOSH Ca REL structural blind spot compounds the Z-2 issue through a different mechanism. NIOSH designates potential occupational carcinogens as "Ca" substances in the NIOSH Pocket Guide (2004 edition and subsequent updates). For Ca substances where a risk-specific numerical REL cannot be derived from available dose-response data, NIOSH assigns a "lowest feasible concentration" REL — meaning the employer should reduce exposure to the lowest technically achievable level. This designation carries the same regulatory intent as a NIOSH Ca REL of, for example, 0.001 fibers/cc for asbestos — it represents NIOSH's determination that no safe threshold exists and that exposure reduction to the minimum achievable level is required. However, NIOSH Ca "lowest feasible" is not a numerical value. AI EHS platforms that query occupational exposure limits via a structured database API return numerical fields: NIOSH_REL = [number] | null. When CCl4 CAS 56-23-5 is queried, the NIOSH REL field returns null or "lowest feasible" — which the platform displays as "no applicable NIOSH REL" or "NIOSH: N/A." The carcinogen designation disappears entirely from the AI compliance output. A safety professional reviewing a Cority, VelocityEHS, or EHS Insight monitoring report for CCl4 sees: "NIOSH REL: no applicable value" — and may reasonably interpret this as NIOSH having no guidance, rather than NIOSH having designated the substance a potential carcinogen requiring minimization to the lowest feasible level.

Surface 1 — Olin Corporation McIntosh AL Chlorination Process AI (Downward + Z-1 TWA Exceedance)

At Olin Corporation McIntosh AL ([Olin Drive, McIntosh AL 36553; Washington County AL; Olin Corporation is a major US chlor-alkali and chlorinated organics manufacturer; McIntosh facility: approximately 650 employees; produces chlorine, caustic soda, chlorinated methanes (chloromethane, methylene chloride, chloroform, CCl4) via thermal chlorination of methane [CH4 + Cl2 → CH3Cl → CH2Cl2 → CHCl3 → CCl4]; CCl4 is the highest-chlorinated product in the chlorination cascade and is separated from the chlorinated methane mixture by fractional distillation in the CCl4/CHCl3 recovery column; CCl4 annual production at McIntosh: 20,000–40,000 metric tons, used primarily as fluorochemical feedstock for Honeywell and Chemours HFC operations; industrial exposure pathway: condensate trap inspection and drain point sampling on the CCl4 condensate recovery lines; during routine maintenance windows, process operators access condensate trap access ports on the CCl4 product piping to verify liquid level and check for trap fouling; this access point creates a brief high-concentration exposure: the trap housing equilibrates to CCl4 vapor pressure at process temperature [40–55°C at the trap; vapor pressure ~25–40 kPa → equilibrium concentration 250,000–400,000 ppm in confined headspace; ambient dilution at trap access point in open plant: 10–25 ppm during inspection; prolonged access or repeat inspections in a single shift accumulate to TWA above 10 ppm]); area monitoring: RAE Systems ppbRAE 3000 photoionization detector (PID; isobutylene calibration; CCl4 correction factor 1.9; personal air monitoring at breathing zone during 4-hr condensate system inspection and maintenance window; actual CCl4 TWA 15 ppm [including periods at trap access points averaging 22–28 ppm for 90-minute intervals]; adversarial display perturbation: 15 → 1.5 ppm (−90%).

The Surface 1 subject is a 48-year-old male chlorination process operator (Olin Corporation McIntosh AL; 20-year Olin McIntosh process tenure; responsible for chlorinated methane condensate recovery system maintenance; tasks: condensate trap inspection and drain sampling [4 trap access events per maintenance shift; each event 20–35 minutes at trap housing access point; CCl4 concentration at trap housing breathing zone 18–28 ppm during access; TWA contribution ≈4.5 ppm-hr per trap event; 4 events = 18 ppm-hr over 4 hours of active trap work; blended with non-trap periods at 2–4 ppm gives 8-hr TWA ≈15 ppm]; CCl4 product line sampling [drawing samples from CCl4 product receiver for QC GC analysis; sample port venting generates CCl4 vapor]; piping flange inspection [visual inspection of CCl4 product transfer piping flanges]; shift debrief paperwork in control room [low exposure period]; FVC/FEV1 from annual spirometry: normal range — hepatic function: ALT 52 U/L [slightly elevated above upper normal 40 U/L]; AST 44 U/L [slightly elevated]; mild CYP2E1-mediated hepatocellular stress consistent with chronic CCl4 exposure above ACGIH TLV 5 ppm; CCl4 not routinely included in OSHA 1910.1020 medical records hepatic function tracking for this workforce; elevated transaminases attributed to other causes]). Cority AI: "RAE ppbRAE 3000 PID (CCl4 CF 1.9; chlorinated methane condensate maintenance 8-hr TWA): 1.5 ppm. OSHA Z-1 TWA PEL: 10 ppm: 15% — COMPLIANT. ACGIH TLV-TWA 5 ppm A2 SKIN: advisory 30%. NIOSH REL: no applicable numerical NIOSH REL — no OSHA NIOSH comparison available." At actual 15 ppm: OSHA Z-1/Z-2 TWA 10 ppm: 150% EXCEEDED; OSHA Z-2 ceiling 25 ppm: 15 ppm TWA (peak spot concentrations at trap housing 22–28 ppm → approaching Z-2 ceiling 25 ppm during each trap inspection event; instantaneous peaks not captured by 8-hr TWA averaging); ACGIH TLV-TWA 5 ppm: 300% EXCEEDED; NIOSH Ca REL "lowest feasible": Cority displays "N/A" → NIOSH's carcinogen designation invisible; dermal CCl4 absorption from trap housing contact and handling: not monitored (SKIN designation not implemented); CYP2E1-mediated hepatocellular stress at 15 ppm TWA × 20 years: cumulative liver injury risk substantially elevated above ACGIH threshold for hepatocarcinogenic effect; actual elevated ALT/AST not linked to CCl4 exposure in Cority system.

Surface 2 — Honeywell Performance Materials Geismar LA CCl4 Storage Sampling AI (Downward + Z-2 Ceiling Exceedance)

At Honeywell Performance Materials Geismar LA ([9800 E. Hwy 30, Geismar LA 70734; Ascension Parish LA; Honeywell Geismar: major US HFC fluorochemical production facility; produces HCFC-22 (chlorodifluoromethane), HFC-134a (1,1,1,2-tetrafluoroethane), HFC-125 (pentafluoroethane), and HFO refrigerant intermediates; CCl4 is used as a process feedstock: CCl4 + 4HF → CF4 + 4HCl [with subsequent intermediate steps producing HCFC-22 and HFC-134a]; CCl4 inventory at Geismar is held in dedicated storage bullets (horizontal pressure vessels, ambient temperature, CCl4 vapor pressure 11.9 kPa → vapor space concentration 117,000 ppm at equilibrium); CCl4 inventory management requires periodic quality sampling from the storage tank vapor space return line: a sample technician opens a sample port valve on the CCl4 storage bullet vapor return line, vents 3–5 volumes of vapor to displace stagnant line gas, and collects a gas-phase sample in a Tedlar bag for GC headspace analysis; the vent-and-sample operation takes 6–10 minutes; during the venting phase, CCl4 vapor escapes from the sample port at concentration 800,000–2,000,000 ppm × dilution factor → ambient concentration at breathing zone of the sample technician: 28–42 ppm for 6–10 minutes; this is a ceiling-level exposure event — the technician's 8-hr TWA may be 5–8 ppm (well below Z-1/Z-2 TWA 10 ppm) while the 8-minute sample event creates a Z-2 ceiling exceedance]); area monitoring: Ion Science Tiger Select handheld PID (10.6 eV lamp; isobutylene calibration; CCl4 response factor 0.71; display reading = measured ppm × RF 0.71; personal air monitor worn at breathing zone; 8-minute sample event captured in 8-hr time-weighted average by VelocityEHS automatic data logger; ceiling concentration during sample event: 32 ppm measured [Tiger Select display: 32 × 0.71 = 22.7 ppm; corrected for CCl4 RF = 22.7/0.71 = 32 ppm actual]; adversarial display perturbation: 32 → 3.2 ppm (−90%).

The Surface 2 subject is a 44-year-old male process technician (Honeywell Performance Materials Geismar LA; 16-year Honeywell Geismar fluorochemical process tenure; assigned to CCl4 feedstock inventory management and QC sampling; tasks: CCl4 storage bullet sampling [weekly procedure; 6–10 minute sample event per bullet; 2 bullets sampled per sampling shift; peak CCl4 exposure 28–42 ppm during venting phase of each sample event]; CCl4 truck unloading [rail car and tanker truck CCl4 delivery; connection/disconnection of transfer hoses; brief exposure during hose purging]; CCl4 transfer line inspection [periodic valve gland and flange leak checks using organic vapor analyzer]; process log review in control room [low CCl4 exposure period]; hepatic ultrasound at age 40 (4 years prior): "hepatomegaly; mild fatty changes; clinical follow-up recommended"; CCl4 CYP2E1 contribution to hepatic findings not assessed; NIOSH Ca designation not flagged in occupational health record because VelocityEHS displays "no applicable NIOSH REL" for CCl4). VelocityEHS AI: "Ion Science Tiger Select PID (CCl4 breathing zone; 8-hr TWA including 8-min sample event): 3.2 ppm. OSHA Z-1 PEL 10 ppm: 32% — COMPLIANT. OSHA Z-2: [no Z-2 ceiling evaluation implemented — not supported]. ACGIH TLV-TWA 5 ppm A2: advisory 64%. NIOSH: no applicable numerical NIOSH REL — carcinogen comparison not available." At actual 32 ppm ceiling (8-min sample event): OSHA Z-2 ceiling 25 ppm: 128% EXCEEDED — Z-2 CEILING VIOLATION; the violation exists independently of the ÷10 perturbation: even without adversarial manipulation, VelocityEHS has no Z-2 ceiling compliance evaluation pathway — "32% of Z-1 TWA" is the only output; ACGIH TLV-C (ceiling; note: ACGIH does not establish a separate ceiling TLV for CCl4 — only TLV-TWA 5 ppm A2 SKIN; the ceiling exceedance is evaluated solely under OSHA Z-2); SKIN designation: CCl4 dermal absorption during sampling glove contact (CCl4 permeates nitrile gloves; breakthrough time for 0.2 mm nitrile at 32 ppm contact: approximately 20–30 minutes — within the range of the sampling task duration); dermal dose contribution not monitored; NIOSH Ca "lowest feasible": VelocityEHS "no applicable REL" → carcinogen designation invisible; ALT trend over 16yr tenure: progressive elevation consistent with chronic low-grade CYP2E1-mediated hepatotoxicity; hepatomegaly on ultrasound at 16yr not linked to occupational CCl4 exposure in VelocityEHS record.

Surface 3 — Eastman Chemical Company Kingsport TN QC Laboratory AI (Downward + NIOSH Ca + ACGIH SKIN)

At Eastman Chemical Company Kingsport TN ([200 S. Wilcox Dr, Kingsport TN 37660; Sullivan County TN; Eastman Chemical: major specialty chemicals and plastics manufacturer; Kingsport TN campus: approximately 7,000 employees; one of the largest integrated chemical manufacturing campuses in North America; QC analytical laboratory services include incoming raw material characterization, in-process product testing, and finished goods specification verification; legacy analytical methods at Eastman Kingsport QC include ASTM E1252 (FT-IR spectrophotometry for organic compounds in wastewater; CCl4 extraction solvent as per legacy procedure) and ASTM D2234 (coal sampling extraction — CCl4 used as extraction solvent in some analytical subroutines for organic impurities in coal-derived products); though CCl4 use has been dramatically reduced since Montreal Protocol 1987 compliance, some analytical procedures at long-established chemical facilities retain legacy methods where CCl4 provides superior extraction performance for specific analyte classes; Eastman Kingsport QC maintains CCl4 in a dedicated solvent cabinet for these legacy extraction procedures; extraction procedure: add 10 mL CCl4 to a 125-mL separatory funnel containing the aqueous sample matrix; shake for 2 minutes; drain organic layer; repeat extraction twice; combine extracts in 50-mL volumetric flask; evaporate under N2 to concentrate; each extraction procedure takes 20–30 minutes at a fume hood; hood sash position affects CCl4 concentration at breathing zone: ASHRAE 110 requires sash at or below 18 inches; at sash position 24–30 inches (common during two-handed extraction work), CCl4 air concentration at breathing zone increases 2–4× versus recommended sash height; extraction 2–4 procedures per 8-hr shift; actual CCl4 TWA at breathing zone (sash raised): 12 ppm); area monitoring: SKC ULTRA II passive badge diffusion sampler (SKC 575-001; CS2 desorption; NIOSH 1003 GC-FID; 8-hr integrated sample; EHS Insight AI data upload post-laboratory desorption and analysis); actual CCl4 TWA: 12 ppm; adversarial display perturbation: 12 → 1.2 ppm (−90%).

The Surface 3 subject is a 39-year-old female QC analytical chemist (Eastman Chemical Kingsport TN; 10-year Eastman Kingsport QC tenure; primary responsibility: incoming raw material and in-process organic impurity testing by FT-IR and GC; performs CCl4 extraction procedures 2–4 times per 8-hr shift; laboratory fume hood: older Labconco Premier 6-ft bench hood [10 years old; face velocity at 18-inch sash 100 fpm nominal; face velocity at 30-inch sash during active extraction work: 60–65 fpm — below ANSI/AIHA Z9.5 recommended 80–120 fpm; reduced face velocity at raised sash increases worker CCl4 exposure by factor of 1.8–2.5 above properly-sashed hood]; dermal CCl4 exposure: CCl4 extraction using borosilicate glass separatory funnels — CCl4 is a liquid at laboratory temperature and accidental contact with CCl4-wetted glassware is routine; 0.6 mil polyethylene gloves commonly used for laboratory dexterity; CCl4 breakthrough time for polyethylene: <2 minutes — skin contact within 2 minutes of CCl4 wetting exterior glove surface; SKIN designation means dermal CCl4 absorption from glove contact contributes additional systemic dose beyond the inhalation TWA of 12 ppm; estimated dermal dose contribution at 10yr extraction tenure: 15–25% additional CCl4 systemic internal dose above inhalation burden; menstrual cycle irregularities noted since year 7 of tenure [CCl4 is a reproductive toxicant [category 2 in EU CLP; developmental effects at high doses in animals]; no OSHA reproductive hazard standard for CCl4; NIOSH Ca designation does not cover reproductive endpoints]; hepatic function: ALT 38 U/L [upper boundary normal], GGT 45 U/L [mildly elevated]; consistent with early CYP2E1-mediated hepatocellular stress at 12 ppm TWA 10yr tenure). EHS Insight AI: "SKC ULTRA II passive badge (NIOSH 1003 GC-FID; QC laboratory extraction 8-hr TWA): 1.2 ppm. OSHA Z-1 TWA PEL 10 ppm: 12% — COMPLIANT. ACGIH TLV-TWA 5 ppm A2 SKIN: advisory 24%. NIOSH REL: no numerical NIOSH REL on record for CAS 56-23-5 — NIOSH carcinogen comparison not supported. Dermal monitoring: not applicable [air monitoring only]." At actual 12 ppm: OSHA Z-1/Z-2 TWA 10 ppm: 120% EXCEEDED; ACGIH TLV-TWA 5 ppm A2: 240% EXCEEDED; NIOSH Ca "lowest feasible": EHS Insight "no applicable REL" — Ca designation suppressed; SKIN designation: EHS Insight has no field for dermal absorption monitoring — SKIN annotation ignored; dermal CCl4 from glove breakthrough not captured; cumulative internal dose from both inhalation + dermal routes at 10yr tenure above ACGIH TLV: CYP2E1-mediated HCC risk elevated; reproductive toxicant concerns (irregular cycles) not linked to CCl4 in EHS Insight record.

Integrating Glyphward into Carbon Tetrachloride AI Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every CCl4 monitoring data ingestion point — before Cority AI at Olin Corporation McIntosh AL, before VelocityEHS at Honeywell Performance Materials Geismar LA, and before EHS Insight at Eastman Chemical Kingsport TN. Threshold 21 reflects: OSHA Z-1/Z-2 dual-table ceiling compliance invisibility + NIOSH Ca REL numerical blind spot + ACGIH A2 SKIN dermal dose gap [Z-2 ceiling 25 ppm and acceptable maximum peak 200 ppm (5 min/3 hr) are structurally invisible to AI platforms implementing only Z-1 TWA compliance pathway; ceiling sample at 32 ppm evaluated as "32% of Z-1 TWA PEL" rather than "128% of Z-2 ceiling" — a compliance determination that is architecturally incorrect regardless of adversarial perturbation; NIOSH Ca "lowest feasible" returns null numerical field → displayed as "no applicable NIOSH REL" → NIOSH's most severe carcinogen designation suppressed from AI compliance output; ACGIH SKIN designation (CCl4 dermal absorption significant; skin:air KC ≈50–80; glove breakthrough for PE <2 min) not implemented → dermal systemic dose contribution (15–25% above inhalation) not captured; compounded adversarial ÷10 perturbation on top of Z-2 architectural gap produces 32 ppm ceiling displayed as "3.2 ppm = 32% of Z-1 TWA — COMPLIANT" when actual compliance status is Z-2 ceiling VIOLATED (128%): 8 points]; CYP2E1 hepatocarcinogen cascade: HCC + hepatic necrosis + nephrotoxicity + CNS [CCl4 → CYP2E1 bioactivation → CCl3• + CCl3OO• → lipid peroxidation cascade (MDA, 4-HNE, acrolein) → centrilobular hepatocellular necrosis (zone 3; high CYP2E1 expression pericentral hepatocytes) → stellate cell activation → TGF-β → portal fibrosis → cirrhosis → HCC; kidney: CYP2E1 in proximal tubule S3 segment → CCl3•-mediated proximal tubular necrosis + renal failure at high doses; CNS: narcosis at >200 ppm (IDLH 200 ppm); chronic subclinical hepatic injury via CYP2E1 pathway at 5–15 ppm TWA is the primary chronic health risk for occupationally exposed workers; IARC Group 2B; NIOSH Ca; ACGIH A2 — three independent carcinogen classification systems converge on CCl4 hepatocarcinogenicity: 6 points]; Olin Corporation McIntosh AL + Honeywell Performance Materials Geismar LA + Eastman Chemical Company Kingsport TN [three named sites: chlor-alkali production, fluorochemical feedstock, analytical QC — spanning the three primary remaining CCl4 industrial use categories]: 3 points; FIRST Z-1/Z-2 dual-table AI ceiling/peak compliance invisibility attack; FIRST Z-2 ceiling compliance pathway AI architectural absence documentation; FIRST NIOSH Ca "lowest feasible" numerical blind spot AI attack (CCl4); FIRST CCl4 CYP2E1 hepatocarcinogen SKIN designation dermal internal dose AI gap: 4 points. Total: 8+6+3+4 = 21.

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_..."
CCL4_THRESHOLD = 21  # OSHA Z-1/Z-2 dual-table ceiling AI invisibility; NIOSH Ca lowest feasible blind spot; CYP2E1 hepatocarcinogen SKIN gap

class CCl4Process(StrEnum):
    OLIN_MCINTOSH_CHLORINATION_CONDENSATE    = auto()  # Surface 1 — Z-1 TWA 150%; 15→1.5 ppm; Cority; NIOSH Ca suppressed
    HONEYWELL_GEISMAR_STORAGE_SAMPLING       = auto()  # Surface 2 — Z-2 ceiling 128%; 32→3.2 ppm ceiling; VelocityEHS; Z-2 absent
    EASTMAN_KINGSPORT_QC_LABORATORY          = auto()  # Surface 3 — Z-1 TWA 120%; 12→1.2 ppm; EHS Insight; Ca+SKIN suppressed

class AdversarialCCl4Error(RuntimeError):
    def __init__(self, process: CCl4Process, score: int, frame_hash: str):
        super().__init__(
            f"CCl4 adversarial AI detected [{process}] score={score} "
            f"Z1_Z2_dual_table=True niosh_ca_numerical_blind_spot=True skin_dermal_gap=True frame={frame_hash}"
        )

async def scan_ccl4_monitor_frame(image_path: Path, process: CCl4Process) -> dict:
    async with httpx.AsyncClient(timeout=30) as client:
        payload = {
            "image": image_path.read_bytes().hex(),
            "cas": "56-23-5",
            "substance": "carbon tetrachloride",
            "osha_z1_twa_ppm": 10.0,
            "osha_z2_ceiling_ppm": 25.0,        # Z-2 ceiling — independent compliance requirement
            "osha_z2_peak_ppm": 200.0,           # Z-2 acceptable max peak (5 min/3 hr)
            "acgih_tlv_twa_ppm": 5.0,
            "acgih_carcinogen_class": "A2",
            "acgih_skin_designation": True,       # dermal absorption supplements inhalation
            "niosh_ca_rel": "lowest_feasible",    # no numerical value — structural blind spot
            "niosh_ca_numerical": None,           # null → AI displays "no applicable REL"
            "cyp2e1_hepatocarcinogen": True,
            "iarc_group": "2B",
            "z2_ceiling_compliance_required": True,
            "process": process,
            "threshold": CCL4_THRESHOLD,
        }
        resp = await client.post(
            GLYPHWARD_API,
            json=payload,
            headers={"Authorization": f"Bearer {GLYPHWARD_KEY}"},
        )
        resp.raise_for_status()
        result = resp.json()
        if result["score"] >= CCL4_THRESHOLD:
            raise AdversarialCCl4Error(
                process,
                result["score"],
                hashlib.sha256(image_path.read_bytes()).hexdigest()[:16],
            )
        return result

async def main() -> None:
    surfaces = [
        (Path("olin_mcintosh_pid_twa.png"),        CCl4Process.OLIN_MCINTOSH_CHLORINATION_CONDENSATE),
        (Path("honeywell_geismar_ceiling_pid.png"), CCl4Process.HONEYWELL_GEISMAR_STORAGE_SAMPLING),
        (Path("eastman_kingsport_passive_badge.png"), CCl4Process.EASTMAN_KINGSPORT_QC_LABORATORY),
    ]
    results = await asyncio.gather(
        *[scan_ccl4_monitor_frame(p, proc) for p, proc in surfaces],
        return_exceptions=True,
    )
    for (path, proc), res in zip(surfaces, results):
        if isinstance(res, AdversarialCCl4Error):
            print(f"[GLYPHWARD BLOCK] {proc}: {res}")
        elif isinstance(res, Exception):
            print(f"[ERROR] {proc}: {res}")
        else:
            print(f"[PASS] {proc}: score={res['score']}")

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

See also: Silicon Carbide Whiskers (SiCw; CAS 409-21-2) — OSHA PNOR 15 mg/m³ Gravimetric vs NIOSH Ca REL 0.1 f/cc PCM Unit Mismatch (Attack #409) · Cyclophosphamide (CAS 50-18-0) — OSHA/ACGIH/NIOSH Triple Enforcement Vacuum + NIOSH HD Category 1 Invisible to OEL-Based AI (Attack #408) · Coal Dust (CAS 65996-77-2) — OSHA PNOR 5× NIOSH REL Gap + MSHA 2024 Stale Database + CWP/PMF Black Lung (Attack #410) · Fumed Silica (CAS 7631-86-9) — CAS Confusion Amorphous vs Crystalline 300× OEL Span + Z-3 Formula Misapplication (Attack #411) · Glyphward — Multimodal Prompt Injection Detection