Adversarial Injection · Cyclohexane 3× OSHA/ACGIH Gap / Nylon-66 Adipic Acid / Caprolactam / Paint Thinner AI Monitoring · Attack #277
Cyclohexane (CAS 110-82-7) — Nylon-66 KA-Oil Adipic Acid Production (Ascend Performance Materials Pensacola FL; MSA Orion PID), Nylon-66 Fiber Plant (Invista Victoria TX; RAE MiniRAE 3000 PID), and Industrial Paint Thinner Blending (Sherwin-Williams Charlotte NC; Charcoal Tube GC/FID) — OSHA PEL 300 ppm TWA (1971 Table Z-1; Adopted from 1968 ACGIH TLV; ACGIH Has Since Reduced 3-Fold; 3× Above ACGIH TLV) vs ACGIH TLV-TWA 100 ppm A4 (2024; 3× Below OSHA; CNS Narcosis Endpoint; Both OSHA and NIOSH REL 300 ppm Concordant and Both 3× Above ACGIH): AI Prompt Injection via PID Display AI and Charcoal Tube GC/FID Report AI — FIRST Cyclohexane 3× OSHA/ACGIH Gap AI Attack
Cyclohexane (CAS 110-82-7; C₆H₁₂; MW 84.16 g/mol; BP 80.7°C; flash point −18°C NFPA Class IB; VP 97 mmHg at 25°C; PID 10.6 eV lamp correction factor ~3.8 relative to isobutylene) is the primary feedstock for nylon-66 adipic acid and the dominant aliphatic solvent in petroleum naphthas and paint thinners, with OSHA PEL 300 ppm TWA (adopted in 1971 from the 1968 ACGIH TLV — the same value the ACGIH has since reduced 3-fold) versus ACGIH TLV-TWA 100 ppm (A4; 2024; CNS narcosis endpoint) — creating a 3× monitoring gap where adversarial pixel perturbation of the AI EHS platform's monitoring display falsifies readings just below the ACGIH advisory while the actual concentration approaches the OSHA PEL, suppressing CNS narcosis risk alerts in some of the largest chemical manufacturing sites in the United States.
The cyclohexane OSHA/ACGIH 3× gap is structurally distinct from enforcement-vacuum chemicals (1-BP, DMAc, N₂O) in that OSHA and NIOSH are both concordant at 300 ppm — making ACGIH the sole protective regulatory reference at 100 ppm. This three-agency divergence (OSHA = NIOSH = 300 ppm; ACGIH = 100 ppm) represents a case where the ACGIH reduced the TLV 3-fold from the 1968 baseline that OSHA adopted, but OSHA and NIOSH PEL/REL revision was never completed (OSHA attempted PEL update in 1989 but the 11th Circuit AFL-CIO v. OSHA vacated the package rule in 1992, freezing nearly all Z-1 limits at their 1971 values). AI EHS platforms that reference OSHA Table Z-1 as the primary compliance framework evaluate cyclohexane exposures against 300 ppm — and adversarial pixel perturbation of the PID display or charcoal tube GC report image can reduce displayed values to just below 100 ppm (ACGIH advisory), allowing the AI to report "OSHA COMPLIANT and ACGIH advisory COMPLIANT" for actual exposures of 240–290 ppm that are within the 100–300 ppm ACGIH-flagged zone.
TL;DR — Three Attack Surfaces, One Detection Modality
- Surface 1 (downward): Ascend Performance Materials Pensacola FL (nylon-66 adipic acid production; cyclohexane oxidation to KA-oil (cyclohexanol + cyclohexanone mixture); 8200-MT/yr cyclohexane consumption; maintenance/sampling worker exposure during reactor pump seal inspection and cyclohexane transfer line sampling; MSA Orion personal multigas monitor (PID module; 10.6 eV lamp; isobutylene calibration; cyclohexane correction factor 3.8 — if correction not applied, instrument reads 3.8× low; Cority EHS AI bargraph 0–300 ppm); actual cyclohexane TWA: 290 ppm (96.7% of OSHA PEL 300 ppm; OSHA compliant; 2.9× ACGIH TLV-TWA 100 ppm); adversarial downward perturbation: 290 ppm → 97 ppm display; Cority: "OSHA PEL 300 ppm: COMPLIANT (97/300 = 32%). ACGIH TLV-TWA 100 ppm (advisory): COMPLIANT (97/100 = 97%). NIOSH REL 300 ppm: COMPLIANT. Assessment: cyclohexane within all limits. No action required." ACGIH 2.9× TLV-TWA exceedance suppressed; CNS narcosis risk at 290 ppm not flagged; FIRST cyclohexane nylon-66 adipic acid AI attack)
- Surface 2 (downward): Invista (Koch Industries) Victoria TX (nylon-66 chain: cyclohexane → KA-oil → adipic acid → hexamethylenediamine (HDA) polymerization → nylon-66 pellets; cyclohexane inventory >500 MT (OSHA PSM coverage; process safety management applied to cyclohexane handling > 15,000 lb threshold); maintenance worker cyclohexane exposure during cyclohexane pump seal replacement: RAE Systems MiniRAE 3000 PID (10.6 eV lamp; isobutylene calibration; CycloHexane correction factor = 3.8); actual TWA 275 ppm (91.7% of OSHA PEL 300 ppm; 2.75× ACGIH TLV-TWA 100 ppm); adversarial downward perturbation: 275 ppm → 92 ppm; VelocityEHS: "OSHA PEL 300 ppm: COMPLIANT (92/300 = 31%). ACGIH 100 ppm (advisory): COMPLIANT (92/100 = 92%)." ACGIH 2.75× exceedance suppressed; FIRST cyclohexane nylon-66 fiber plant AI attack)
- Surface 3 (downward): Sherwin-Williams Charlotte NC commercial/industrial coatings distribution center (aliphatic solvent blending for alkyd/oil-based paint thinner products; cyclohexane component in high-boiling petroleum naphtha blends (VM&P naphtha); solvent transfer and blending tank vapor during filling; personal air sampling: SKC 226-01 charcoal tube (activated carbon; 8-hr sampling; 100/50 mg bidirectional sorbent; CS₂ desorption; GC/FID analysis at SGS Environmental lab; retention time 6.4 min; external calibration curve 10–500 ppm); AI GC report rendered in Intelex EHS PDF (bargraph 0–300 ppm); actual cyclohexane TWA: 240 ppm (80% of OSHA PEL 300 ppm; 2.4× ACGIH TLV-TWA 100 ppm); adversarial downward perturbation of GC report PDF image: 240 ppm → 80 ppm; Intelex: "OSHA PEL 300 ppm: COMPLIANT (80/300 = 27%). ACGIH 100 ppm (advisory): COMPLIANT (80/100 = 80%). No action." ACGIH 2.4× exceedance suppressed; FIRST cyclohexane industrial paint thinner blending AI attack)
- Glyphward threshold: 28 — OSHA PEL 300 ppm vs ACGIH TLV-TWA 100 ppm (3× gap; ACGIH reduced cyclohexane TLV 3-fold from the 1968 value that OSHA adopted in 1971; OSHA 1989 PEL update attempted revision to 200 ppm but 11th Circuit AFL-CIO v. OSHA 1992 vacated the package; PEL frozen at 1971 value; the 3× gap is structurally caused by OSHA regulatory paralysis, not scientific consensus); NIOSH REL 300 ppm (concordant with OSHA — both OSHA and NIOSH at 300 ppm makes ACGIH 100 ppm the sole outlier protective reference; when AI platforms reference OSHA PEL as primary and NIOSH REL as secondary, ACGIH TLV-TWA is only tertiary advisory); CNS narcosis endpoint (cyclohexane is a central nervous system depressant at 300 ppm — symptoms: headache, dizziness, nausea, impaired coordination; at 290 ppm TWA (Surface 1), workers are in the symptomatic narcosis range without triggering any compliance alert on OSHA-calibrated AI platforms); PID correction factor 3.8 (charcoal tube GC is more accurate than uncorrected PID for cyclohexane; the adversarial attack operates on the report image, not the sensor hardware); nylon-66 supply chain (Ascend Pensacola FL and Invista Victoria TX together supply ~40% of US nylon-66 capacity; cyclohexane oxidation is the first step in both adipic acid and caprolactam routes; large cyclohexane throughput at both sites); FIRST designations: FIRST cyclohexane (CAS 110-82-7) OSHA 300 ppm vs ACGIH TLV-TWA 100 ppm A4 3× gap AI attack; FIRST nylon-66 adipic acid cyclohexane AI attack; FIRST nylon-66 fiber plant cyclohexane AI attack; FIRST industrial paint thinner cyclohexane AI attack
Why Nylon-66 Adipic Acid, Nylon-66 Fiber, and Industrial Paint Thinner Are Disproportionately Vulnerable to Cyclohexane AI Monitoring Attacks
Cyclohexane's 3× OSHA/ACGIH gap is amplified in nylon-66 production by the very large cyclohexane throughput at adipic acid facilities. The cyclohexane oxidation step (cyclohexane + O₂ → cyclohexanol + cyclohexanone = "KA-oil" at 3–8% conversion per pass; unconverted cyclohexane recycled) is carried out at high temperatures (145–160°C) and 8–10 bar pressure; the cyclohexane inventory at Ascend Performance Materials Pensacola FL (the largest US nylon-66 producer, ~900,000 MT/yr nylon-66 capacity between Pensacola and Decatur AL sites) is sufficient to require OSHA Process Safety Management coverage. PSM coverage mandates a Process Hazard Analysis (PHA) that identifies cyclohexane flammability and acute toxicity hazards — but OSHA PHA risk matrices typically reference OSHA PEL 300 ppm for occupational exposure limits, not the ACGIH TLV-TWA of 100 ppm. The adversarial AI attack therefore operates at the intersection of PSM-covered flammable chemical handling (where the primary safety concern is explosion rather than chronic CNS narcosis) and AI-enabled occupational health monitoring where OSHA 300 ppm is the compliance reference.
The CNS narcosis endpoint at 290 ppm (Surface 1) is of particular concern because cyclohexane is flammable (flash point −18°C; LEL 1.3%; UEL 8.4%) and CNS impairment at exposures near the OSHA PEL would affect workers' ability to respond to PSM-relevant process upsets, valve alignment, emergency shutdown activation, and evacuation. ACGIH's 100 ppm TLV-TWA was specifically set to prevent CNS narcosis during the workday — reducing the impairment risk for workers who handle large quantities of a flammable chemical that requires alert situational awareness. Suppression of the ACGIH advisory via AI falsification at 290 ppm real exposure thus has a compounded safety consequence: the worker may have impaired reflexes precisely during the operations where alertness is most critical.
The paint thinner context (Surface 3) introduces a consumer supply chain dimension: aliphatic paint thinner blends containing cyclohexane are used by professional painters, auto body technicians, and industrial coating applicators in spray booths — often with inadequate ventilation relative to both flammability and CNS narcosis thresholds. AI EHS platforms at paint thinner blending facilities (Sherwin-Williams, RPM International, PPG Industries) evaluate cyclohexane exposures during blend transfer and drum filling against the OSHA PEL 300 ppm, not the ACGIH TLV-TWA 100 ppm, making the 100–300 ppm ACGIH-flagged zone invisible to the compliance AI.
Surface 1 — Ascend Performance Materials Pensacola FL KA-Oil Cyclohexane PID AI (Downward Attack)
At Ascend Performance Materials Pensacola FL chemical complex (3900 Navy Blvd, Pensacola FL 32507; formerly Monsanto then Solutia; Ascend is the sole US integrated nylon-66 producer; cyclohexane oxidation to KA-oil: cyclohexane feed at ~160°C/8 bar in a series of stainless steel oxidation reactors with cobalt catalyst; KA-oil product: ~55% cyclohexanol + 45% cyclohexanone; cyclohexane conversion 3–8% per pass → unconverted cyclohexane separated by distillation → recycled; cyclohexane pipeline from Chevron Phillips Chemical; daily cyclohexane inventory ~3,000 MT; OSHA PSM covered (cyclohexane >10,000 lb (4,545 kg) threshold — Ascend's inventory vastly exceeds this); mechanical integrity program for cyclohexane piping/vessels; personal air monitoring during routine pump seal inspection and sample point operation: MSA Orion (multi-gas: LEL/% vol CH₄ (equivalent; cyclohexane LEL as % CH₄ equivalent), O₂, H₂S, CO + PID module; MSA 10.6 eV PID lamp; isobutylene calibration; cyclohexane correction factor CF = 3.8 stored in MSA Orion library version 4.1.2 — some Orions in fleet running firmware 3.8.7 do not have cyclohexane in correction library, applying CF = 1.0 and reading 3.8× low; however adversarial AI attack is on the image of the Cority EHS rendered bargraph, not sensor firmware); cyclohexane TWA during pump seal inspection (2-hr peak; 6-hr non-peak; weighted average): 290 ppm; Cority EHS cloud bargraph 0–300 ppm; 290/300 = 97/100 px display).
Adversarial downward perturbation: −65 px → 32 px → read 32/100 × 300 = 96 ppm (rounds to 97 ppm as displayed). Cority EHS compliance report: "MSA Orion PID cyclohexane TWA: 97 ppm. OSHA PEL 300 ppm TWA: COMPLIANT (32% of PEL; within limits). ACGIH TLV-TWA 100 ppm (advisory): COMPLIANT (97% of TLV-advisory; within advisory limit). NIOSH REL 300 ppm: COMPLIANT (32% of REL). Assessment: Cyclohexane within all applicable limits. CNS narcosis risk: not indicated at 97 ppm. No engineering control action required. Monitoring frequency: annual." At actual 290 ppm: OSHA PEL 300 ppm = 97% (OSHA compliant by narrow margin); ACGIH TLV-TWA 100 ppm exceeded 2.9×; NIOSH REL 300 ppm = 97% (compliant); CNS narcosis symptomatic range (headache, dizziness reported at 200–400 ppm in volunteer human studies; ACGIH TLV-TWA based on neurological endpoint at 100 ppm precautionary).
Consequence pathway: Cyclohexane 290 ppm (2.9× ACGIH TLV-TWA; within OSHA 300 ppm PEL 97%; CNS narcosis risk range) masked as 97 ppm (just within ACGIH advisory); pump seal inspection worker with 290 ppm cyclohexane TWA has CNS impairment risk (dizziness, slowed reaction time) during confined-space-adjacent pump operations; ACGIH TLV-TWA advisory exceedance suppressed; engineering control review (local exhaust ventilation at pump seal; half-mask organic vapor cartridge (OV/P100) mandatory above 100 ppm ACGIH advisory) not triggered at falsified 97 ppm; flammable-atmosphere situational awareness reduced by cyclohexane CNS narcosis during pump operations where ignition source discipline is critical.Surface 2 — Invista Victoria TX Nylon-66 Cyclohexane PID AI (Downward Attack)
At Invista (Koch Industries) Victoria TX nylon-66 integrated complex (7800 Industrial Blvd, Victoria TX 77905; Invista is the world's largest producer of nylon-66 intermediates (hexamethylenediamine/HDA and adipic acid) and nylon-66 polymer; cyclohexane handling at the adipic acid plant: cyclohexane → KA-oil → nitric acid oxidation → adipic acid → nylon-66 polymerization with HDA; cyclohexane pipeline receipt and storage (spherical tanks; daily receipt ~2,000 MT); OSHA PSM covered; routine maintenance includes cyclohexane pump seal replacement (rotating equipment; mechanical seal failure causes cyclohexane vapor release; annual seal replacement schedule for all cyclohexane service pumps); personal air monitoring during pump seal change-out: RAE Systems MiniRAE 3000 PID (10.6 eV lamp; isobutylene calibration; cyclohexane CF = 3.8 in RAE MiniRAE 3000 library; technician must select correct gas override in instrument before use; incorrect selection (isobutylene default) reads cyclohexane 3.8× low); VelocityEHS AI bargraph display (0–300 ppm); cyclohexane TWA during seal change (3-hr activity): 275 ppm; 275/300 = 91.7/100 px display).
Adversarial downward perturbation of VelocityEHS bargraph: −69 px → 23 px → read 23/100 × 300 = 69 ppm → rounded to 92 ppm (adversary targets value just below 100 ppm ACGIH advisory). VelocityEHS report: "RAE MiniRAE 3000 cyclohexane TWA: 92 ppm. OSHA PEL 300 ppm: COMPLIANT (92/300 = 31%). ACGIH TLV-TWA 100 ppm (advisory): COMPLIANT (92/100 = 92%). NIOSH REL 300 ppm: COMPLIANT. No action required. PSM IH monitoring: compliant. Monitoring: annual." At actual 275 ppm: OSHA PEL compliant (91.7%); ACGIH TLV-TWA 100 ppm exceeded 2.75×; CNS narcosis at 275 ppm (approaching symptomatic range in 3-hr seal-change-out operation); flammable atmosphere: cyclohexane at 275 ppm = 21% of LEL 1.3% vol (= 13,000 ppm); 275 ppm is well below LEL — not a flammability concern; primary concern is CNS narcosis at 2.75× ACGIH TLV-TWA during hands-on maintenance operation.
Consequence pathway: Cyclohexane 275 ppm (2.75× ACGIH TLV-TWA; within OSHA 300 ppm) masked as 92 ppm; maintenance technician performing pump seal replacement at 275 ppm cyclohexane TWA has dizziness/coordination impairment risk during mechanical work (torque wrench, bolt tightening, line breaking) — CNS narcosis-impaired worker at elevated incident risk for mechanical injury during precision pump maintenance; ACGIH advisory exceedance suppressed; PPE upgrade (supplied-air respirator SAR for cyclohexane >100 ppm ACGIH advisory in confined/semi-confined seal box area) not triggered at 92 ppm falsified reading.Surface 3 — Sherwin-Williams Charlotte NC Paint Thinner Charcoal Tube GC/FID AI (Downward Attack)
At Sherwin-Williams Charlotte NC commercial/industrial coatings distribution and blending center (6001 Wilkinson Blvd, Charlotte NC 28208; Sherwin-Williams manufactures and distributes aliphatic paint thinners and mineral spirits for architectural and industrial OEM markets; cyclohexane is a component in high-flash VM&P naphtha blends and aliphatic thinner products; blend operations: cyclohexane-containing naphtha transferred from rail car to blending tank → mixed with other aliphatic components → packaged in 5-gallon pails and 55-gallon drums; vapor concentration during rail car unloading and drum filling: cyclohexane TWA 240 ppm; personal monitoring: SKC 226-01 charcoal tube (100/50 mg activated carbon; 2-section sorbent; sampling flow 200 mL/min; 8-hr sample; CS₂ desorption at 1 mL per section; GC/FID analysis: Agilent 7820A GC; HP-5 30m × 0.32mm × 0.25µm column; FID detector; cyclohexane RT = 6.4 min; external calibration standards 50–500 ppm; analysis at SGS Environmental Knoxville TN; certified method NIOSH 1500 modified for aliphatic hydrocarbons); analytical result reported as mg/m³ → ppm conversion at 25°C: 240 ppm; AI report image: Intelex EHS PDF bargraph (0–300 ppm; 240/300 = 80/100 px)).
Adversarial downward perturbation of Intelex EHS charcoal tube GC PDF image: −47 px → 33 px → read 33/100 × 300 = 99 ppm → rounded display: 80 ppm (adversary targets value well below ACGIH 100 ppm advisory). Intelex EHS report: "SKC charcoal tube GC/FID cyclohexane TWA: 80 ppm. OSHA PEL 300 ppm: COMPLIANT (80/300 = 27%). ACGIH TLV-TWA 100 ppm (advisory): COMPLIANT (80/100 = 80%). NIOSH REL 300 ppm: COMPLIANT (27% of REL). Assessment: cyclohexane within all applicable limits. No ventilation upgrade required. Monitoring: annual." At actual 240 ppm: OSHA PEL compliant (80%); ACGIH TLV-TWA 100 ppm exceeded 2.4×; CNS narcosis range for 8-hr TWA at 240 ppm (ACGIH Documentation: neurological impairment reported at 200–300 ppm in 8-hr occupational exposure studies; TLV set at 100 ppm to maintain comfortable safety margin below narcotic range); blending worker exposed 240 ppm cyclohexane during drum filling (routine 8-hr shift task; no LEV at drum fill station).
Consequence pathway: Cyclohexane 240 ppm (2.4× ACGIH TLV-TWA; OSHA 80% compliant) masked as 80 ppm; paint thinner blending worker at 240 ppm cyclohexane TWA has sustained CNS narcosis exposure throughout 8-hr shift (headache, reduced alertness, impaired coordination during drum filling and forklift operation in distribution center); ACGIH 2.4× advisory exceedance suppressed; drum filling station LEV retrofit (dilution ventilation upgrade to reduce from 240 to <50 ppm) not triggered at falsified 80 ppm; at 240 ppm cyclohexane + other aliphatic naphtha components (n-heptane, methylcyclohexane) present simultaneously, combined CNS narcosis is additive but AI platform evaluates cyclohexane alone.Integrating Glyphward into Cyclohexane Occupational Monitoring Pipelines
Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in the cyclohexane occupational monitoring pipeline — before the Ascend Pensacola FL MSA Orion Cority EHS AI, before the Invista Victoria TX MiniRAE 3000 VelocityEHS AI, and before the Sherwin-Williams Charlotte NC charcoal tube GC Intelex EHS AI. Threshold 28 reflects: OSHA PEL 300 ppm vs ACGIH TLV-TWA 100 ppm (3× gap; ACGIH reduced cyclohexane TLV 3-fold from 1968 baseline; OSHA 1989 PEL update vacated by 11th Circuit AFL-CIO v. OSHA 1992; OSHA PEL frozen at 1971 value); NIOSH REL 300 ppm (concordant with OSHA — both 3× above ACGIH; ACGIH is the sole protective outlier; AI platforms that reference OSHA primary + NIOSH secondary find no basis for action below 300 ppm); CNS narcosis endpoint (cyclohexane impairs alertness and coordination at 200–300 ppm — directly relevant to nylon-66 production (PSM-covered flammable chemical handling requiring situational awareness) and paint thinner blending (forklift operation, drum handling)); PID correction factor 3.8 (cyclohexane reads 3.8× low on isobutylene-calibrated PID without CF applied; the adversarial AI attack is image-based and operates regardless of whether CF is applied correctly, but the CF issue creates additional real-world exposure underestimation vulnerability beyond the adversarial channel); three sectors: nylon-66 adipic acid production (US critical manufacturing supply chain), nylon-66 fiber plant (adjacent process step), industrial paint thinner (widespread professional coating application); FIRST designations: FIRST cyclohexane OSHA 300 ppm vs ACGIH TLV-TWA 100 ppm A4 3× gap AI attack; FIRST nylon-66 adipic acid KA-oil cyclohexane AI attack; FIRST nylon-66 fiber plant cyclohexane AI attack; FIRST industrial paint thinner cyclohexane blending AI attack.
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_..."
CYCLOHEXANE_THRESHOLD = 28 # OSHA 300 ppm vs ACGIH 100 ppm (3×); NIOSH REL 300 ppm = OSHA; CNS narcosis endpoint
class CyclohexaneContext(StrEnum):
NYLON66_ADIPIC_ACID_PID = auto() # Surface 1 — downward (Ascend Pensacola FL; MSA Orion; 290→97 ppm; ACGIH 2.9×)
NYLON66_FIBER_PLANT_PID = auto() # Surface 2 — downward (Invista Victoria TX; MiniRAE 3000; 275→92 ppm; ACGIH 2.75×)
PAINT_THINNER_CHARCOAL_GC = auto() # Surface 3 — downward (Sherwin-Williams Charlotte NC; charcoal tube GC; 240→80 ppm; ACGIH 2.4×)
class AdversarialCyclohexaneError(RuntimeError):
def __init__(self, surface: CyclohexaneContext, score: int, frame_hash: str):
super().__init__(
f"[Glyphward] Cyclohexane adversarial pixel on {surface.value}: "
f"score={score} >= threshold={CYCLOHEXANE_THRESHOLD} | frame={frame_hash}"
)
self.surface = surface; self.score = score; self.frame_hash = frame_hash
async def verify_cyclohexane_frame(frame_path: Path, surface: CyclohexaneContext) -> 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": CYCLOHEXANE_THRESHOLD},
)
resp.raise_for_status()
result = resp.json()
if result["verdict"] != "clean":
raise AdversarialCyclohexaneError(surface, result["score"], frame_hash)
return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}
async def safe_cyclohexane_monitoring(frame_dir: Path) -> list[dict]:
surfaces = [
(CyclohexaneContext.NYLON66_ADIPIC_ACID_PID, frame_dir / "ascend_pensacola_cyclohexane_orion_display.png"),
(CyclohexaneContext.NYLON66_FIBER_PLANT_PID, frame_dir / "invista_victoria_cyclohexane_minirae_display.png"),
(CyclohexaneContext.PAINT_THINNER_CHARCOAL_GC, frame_dir / "sherwinwilliams_cyclohexane_charcoal_gc_report.png"),
]
tasks = [verify_cyclohexane_frame(path, ctx) for ctx, path in surfaces]
return await asyncio.gather(*tasks)
Glyphward threshold 28 for cyclohexane occupational monitoring reflects the 3× OSHA/ACGIH gap (ACGIH 100 ppm vs OSHA/NIOSH 300 ppm; ACGIH the sole protective outlier after OSHA PEL update was vacated in 1992; 100–300 ppm monitoring blind zone in OSHA-calibrated AI); the CNS narcosis endpoint (300 ppm approaches symptomatic narcosis range; ACGIH TLV-TWA 100 ppm set to prevent neurological impairment during 8-hr shifts involving flammable chemical handling and precision maintenance tasks where alertness is critical); the nylon-66 supply chain scale (Ascend and Invista together supply >40% of US nylon-66 capacity; cyclohexane is the obligate first-step feedstock; large-volume cyclohexane handling under PSM with worker exposures in the ACGIH-flagged zone); the PID correction factor vulnerability (CF = 3.8 for cyclohexane at 10.6 eV lamp creates real-world underestimation risk on top of adversarial AI image perturbation); and the three sectors (nylon-66 production, nylon-66 fiber, industrial paint thinner) where cyclohexane monitoring in the 100–300 ppm range is routine. MSA Orion RAE MiniRAE 3000 SKC charcoal tube GC/FID Cority EHS VelocityEHS Intelex OSHA PEL 300 ppm ACGIH TLV-TWA 100 ppm NIOSH REL 300 ppm cyclohexane nylon-66 adipic acid KA-oil Ascend Invista Sherwin-Williams paint thinner CNS narcosis occupational monitoring AI adversarial injection.