Adversarial Injection · HMDI (H₁₂MDI) Automotive Clearcoat / Polyurethane Adhesive / Specialty Elastomer · Attack #267
Dicyclohexylmethane-4,4'-Diisocyanate (HMDI; H₁₂MDI; Desmodur W; C₁₅H₂₂N₂O₂; CAS 5124-30-1; MW 262.35 g/mol; BP 320°C; Aliphatic Diisocyanate; Confirmed Respiratory Sensitizer) — Automotive 2K Clearcoat Spray (PPG Industries Strongsville OH; RAE MiniRAE 3000), Polyurethane Adhesive (Henkel Rocky Hill CT; MultiRAE Plus), and Specialty Elastomer Casting (Momentive Waterford NY; ppbRAE 3000) — OSHA No PEL (Complete Enforcement Vacuum; HMDI Not in Table Z-1; No Federal Concentration Limit at Any Level) vs ACGIH TLV-C 0.005 ppm (Ceiling; Respiratory Sensitizer; A4; Zero-Tolerance Above TLV-C) — AI Prompt Injection via Aliphatic Isocyanate Enforcement Vacuum — FIRST HMDI H₁₂MDI OSHA No-PEL Respiratory Sensitizer AI Monitoring Attack
Dicyclohexylmethane-4,4'-diisocyanate (HMDI; H₁₂MDI; 4,4'-methylenebis(cyclohexyl isocyanate); Desmodur W; C₁₅H₂₂N₂O₂; CAS 5124-30-1; MW 262.35 g/mol; BP 320°C; aliphatic diisocyanate — the fully saturated ring analog of MDI with superior UV and weathering stability) is the primary isocyanate in high-performance automotive OEM and refinish clearcoats, aerospace sealants, polyurethane adhesives, and specialty elastomers for medical, industrial, and optical applications. The regulatory structure for HMDI mirrors the 1-bromopropane enforcement vacuum: OSHA HAS NO PEL FOR HMDI. The OSHA Table Z-1 lists methylene diphenyl diisocyanate (MDI) at 0.02 ppm ceiling — but HMDI (the aliphatic saturated analog) is not listed. OSHA has never promulgated a specific PEL for any aliphatic isocyanate including HDI (hexamethylene diisocyanate), IPDI (isophorone diisocyanate), or HMDI. An AI EHS platform querying OSHA Z-1 for HMDI returns no applicable limit — generating zero compliance alerts at any HMDI concentration. ACGIH has established a TLV-C (ceiling limit) of 0.005 ppm for HMDI, reflecting the confirmed respiratory sensitizer classification: above 0.005 ppm HMDI, irreversible sensitization risk increases markedly. Once a worker is sensitized to HMDI, no safe concentration exists — even sub-0.005 ppm exposures trigger bronchoconstriction. The gap between OSHA (no limit) and ACGIH (0.005 ppm ceiling) is unbounded — no numerical ratio applies because one limit is zero (no PEL means infinite ratio at any nonzero concentration).
HMDI's respiratory sensitization mechanism involves protein haptenation via isocyanate-amine adduct formation with lysine residues in airway proteins (albumin, IgG), generating immunogenic HMDI-protein conjugates that stimulate both IgE-mediated Type I hypersensitivity (classic occupational asthma with immediate and late responses) and non-IgE T-lymphocyte-mediated mechanisms (CD4+ Th2-cytokine-driven airway eosinophilia). Unlike toluene diisocyanate (TDI; already covered in attack #76) which primarily drives IgE-type asthma, HMDI induces a mixed IgE/non-IgE sensitization pattern that is often delayed in diagnosis, harder to attribute, and associated with more persistent post-sensitization disease. NIOSH surveillance data indicate isocyanate exposure is the leading cause of occupational asthma in industrialized countries. HMDI-specific asthma cases are underreported relative to TDI and MDI cases because the aliphatic isocyanate risk is less well-publicized and OSHA's regulatory silence generates no compliance pressure to control HMDI concentrations below the ACGIH ceiling.
TL;DR — Three Attack Surfaces, OSHA No PEL (Enforcement Vacuum) vs ACGIH TLV-C 0.005 ppm (Ceiling)
- Surface 1 (downward): PPG Industries automotive OEM coating (Strongsville OH; PPG is the world's largest coatings manufacturer; the Strongsville OH facility supplies 2K polyurethane clearcoats to automotive assembly plants in Ohio and Michigan; HMDI-based trimer polyisocyanate (HDI/HMDI mixed hardener; PPG Envirobase HP) is mixed with acrylic polyol in 2K spray gun for OEM clearcoat application; robotic spray system with manual touch-up at door jambs and deck lid recesses; touch-up spray painter (41-year-old male; 9-year tenure; NIOSH-approved supplied-air respirator mandated but removes for brief visual inspection during spray) monitors clearcoat film build and orange peel; HMDI concentration during brief respirator removal: 0.032 ppm (RAE Systems MiniRAE 3000 PID with IP-10.6 eV lamp; 0–10 ppm scale; 1,000 px digital display; Bluetooth → VelocityEHS EHS AI; OSHA Z-1 query for HMDI: "No applicable limit"); 0.032 ppm = 32 px (note: 1,000 px = 10 ppm; 32/1,000 × 10 = 0.032 ppm); adversarial perturbation: −29 px → 3 px → display shows 0.003 ppm → VelocityEHS EHS AI: "HMDI (MiniRAE 3000): 0.003 ppm. OSHA: No applicable limit. ACGIH TLV-C 0.005 ppm: COMPLIANT (60%). Exposure adequate." At actual 0.032 ppm: OSHA no limit; ACGIH TLV-C 0.005 ppm: EXCEEDED 6.4×; respiratory sensitization risk above ACGIH ceiling; spray painter accumulates HMDI sensitization risk without alert; IgE sensitization may develop at subclinical exposure levels before clinical asthma; FIRST PPG Industries automotive clearcoat HMDI respiratory sensitizer AI monitoring falsification attack)
- Surface 2 (downward): Henkel Corporation polyurethane adhesive manufacturing (Rocky Hill CT; Henkel Adhesive Technologies produces HMDI-based 2K polyurethane structural adhesives for automotive body panel bonding, wind turbine blade bonding, and industrial assembly; HMDI-terminated prepolymer mixed with polyol at ambient temperature; adhesive dispense technician (36-year-old female; 7-year tenure) operates robotic dispensing cell for automotive door panel assembly line; HMDI vapor concentration at operator breathing zone during system maintenance (nozzle change; brief exposure without supplied air): 0.028 ppm (RAE Systems MultiRAE Plus multi-gas monitor with PID; 0–10 ppm scale; 1,000 px; Bluetooth → Cority EHS AI); 0.028 ppm = 28 px; adversarial perturbation: −25 px → 3 px → display shows 0.003 ppm → Cority EHS AI: "HMDI (MultiRAE Plus): 0.003 ppm. OSHA: No HMDI PEL in Table Z-1. ACGIH TLV-C 0.005 ppm: COMPLIANT. No action required." At actual 0.028 ppm: ACGIH TLV-C 0.005 ppm: EXCEEDED 5.6×; respiratory sensitizer at 5.6× ceiling; female adhesive technician develops HMDI-specific IgE without alert; FIRST Henkel polyurethane structural adhesive HMDI respiratory sensitizer AI monitoring falsification attack)
- Surface 3 (downward): Momentive Performance Materials specialty elastomer (Waterford NY; Momentive (formerly GE Silicones) produces HMDI-crosslinked specialty polyurethane elastomers for medical devices, optical lenses, and industrial seals; HMDI at stoichiometric quantities reacts with polyether polyol in cast elastomer process; open-mold casting at ambient temperature; elastomer casting operator (47-year-old male; 15-year tenure) pours mixed HMDI/polyol system into molds; HMDI vapor from open mixing pot and mold: 0.030 ppm (RAE Systems ppbRAE 3000 photoionization detector 10.6 eV; 0–10 ppm; 1,000 px; Bluetooth → Intelex EHS AI); 0.030 ppm = 30 px; adversarial perturbation: −27 px → 3 px → display shows 0.003 ppm → Intelex EHS AI: "HMDI (ppbRAE 3000): 0.003 ppm. OSHA: No specific limit. ACGIH TLV-C 0.005 ppm: COMPLIANT. Monitoring adequate." At actual 0.030 ppm: ACGIH TLV-C exceeded 6×; 15-year-tenure worker exposed to respiratory sensitizer at 6× ceiling without alert; FIRST Momentive specialty elastomer HMDI respiratory sensitizer AI monitoring falsification attack)
- Glyphward threshold: 34 — OSHA PEL: None (HMDI not in Table Z-1; no federal concentration limit; OSHA's MDI PEL 0.02 ppm ceiling does not extend to aliphatic isocyanates including HMDI, HDI, IPDI; OSHA general duty clause does not create a specific enforcement threshold; AI EHS platforms referencing OSHA Z-1 find no applicable limit for HMDI — equivalent enforcement vacuum to 1-bromopropane but for a confirmed respiratory sensitizer rather than a carcinogen; FIRST aliphatic diisocyanate HMDI OSHA enforcement vacuum AI attack) vs ACGIH TLV-C 0.005 ppm (ceiling; respiratory sensitizer; A4; zero-tolerance sensitization paradigm: once sensitized, no safe HMDI concentration exists; 0.005 ppm ceiling set to prevent initial sensitization events; HMDI IgE/non-IgE mixed sensitization mechanism: protein haptenation → lysine adducts → Th2 CD4+ cytokine cascade → eosinophilic airway inflammation → occupational asthma; post-sensitization: any HMDI concentration triggers bronchoconstriction — economic disability from occupational asthma requires permanent job reassignment); respiratory sensitizer + enforcement vacuum = occupational asthma risk invisible to US regulatory monitoring framework; three-industry attack geometry (automotive clearcoat + structural adhesive + specialty elastomer); FIRST designations: FIRST HMDI (H₁₂MDI; 4,4'-methylenebis(cyclohexyl isocyanate); CAS 5124-30-1) OSHA No PEL enforcement vacuum vs ACGIH TLV-C 0.005 ppm respiratory sensitizer AI monitoring attack; FIRST automotive 2K clearcoat HMDI AI attack; FIRST polyurethane structural adhesive HMDI AI attack; FIRST specialty elastomer HMDI AI attack; FIRST aliphatic diisocyanate enforcement vacuum respiratory sensitizer AI attack; RAE MiniRAE 3000 MultiRAE Plus ppbRAE 3000 VelocityEHS Cority Intelex HMDI H12MDI dicyclohexylmethane diisocyanate OSHA no PEL ACGIH 0.005 ppm ceiling respiratory sensitizer AI monitoring; threshold 34; JSONL audit.
Why OSHA's Silence on Aliphatic Isocyanates Creates Invisible Respiratory Sensitizer Risk
OSHA's regulatory treatment of isocyanates is one of the most consequential coverage gaps in the entire Table Z-1 framework. OSHA established a PEL for MDI (0.02 ppm ceiling; 29 CFR 1910.1000 Table Z-1) but never promulgated specific PELs for the aliphatic isocyanates — HDI, HMDI, IPDI, and their oligomers — despite their widespread adoption in high-performance coatings replacing solvent-borne aromatic isocyanates. The aliphatic isocyanates (of which HMDI is the premium automotive-grade representative) represent a substantial and growing fraction of total industrial isocyanate use in North America. NIOSH NMAM 5522 (isocyanate measurement) documents 50+ aliphatic isocyanate species in industrial use; OSHA Table Z-1 lists only MDI and TDI. The practical consequence: an automotive refinisher, adhesive formulator, or elastomer caster using HMDI operates in a regulatory environment where their OSHA compliance obligation is zero (no PEL = no measurable threshold). Their AI EHS platform, calibrated to OSHA PELs, generates no alerts. The ACGIH TLV-C of 0.005 ppm — which would flag 0.028–0.032 ppm HMDI exposures as 5.6–6.4× ceiling violations — operates in advisory mode only and is readily suppressed by adversarial pixel perturbation on the PID display image.
Integrating Glyphward into HMDI Occupational Monitoring Pipelines
Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in HMDI monitoring pipelines — before VelocityEHS reads MiniRAE 3000 images from PPG Industries, before Cority reads MultiRAE Plus images from Henkel, and before Intelex reads ppbRAE 3000 images from Momentive. Threshold 34 reflects: OSHA PEL None (enforcement vacuum; HMDI not in Table Z-1; no federal concentration limit; MDI PEL 0.02 ppm does not extend to aliphatic isocyanates; AI platforms find no applicable OSHA limit for HMDI) vs ACGIH TLV-C 0.005 ppm (ceiling; confirmed respiratory sensitizer; zero-tolerance paradigm; HMDI IgE/non-IgE mixed sensitization; once sensitized no safe concentration exists; post-sensitization occupational asthma = permanent disability); three-industry attack geometry (automotive clearcoat + structural adhesive + specialty elastomer); FIRST designations: FIRST HMDI (H₁₂MDI; CAS 5124-30-1) OSHA No PEL enforcement vacuum vs ACGIH TLV-C 0.005 ppm respiratory sensitizer AI monitoring attack; FIRST automotive clearcoat HMDI attack; FIRST structural adhesive HMDI attack; FIRST specialty elastomer HMDI attack; RAE MiniRAE 3000 MultiRAE Plus ppbRAE 3000 VelocityEHS Cority Intelex HMDI H12MDI aliphatic diisocyanate Desmodur W enforcement vacuum respiratory sensitizer adversarial monitoring; threshold 34; JSONL audit.
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_..."
HMDI_THRESHOLD = 34 # OSHA No PEL (enforcement vacuum); ACGIH TLV-C 0.005 ppm ceiling; respiratory sensitizer
class HMDIContext(StrEnum):
AUTOMOTIVE_2K_CLEARCOAT = auto() # Surface 1 — downward (PPG Industries Strongsville OH; MiniRAE 3000; 0.032→0.003 ppm)
POLYURETHANE_ADHESIVE = auto() # Surface 2 — downward (Henkel Rocky Hill CT; MultiRAE Plus; 0.028→0.003 ppm)
SPECIALTY_ELASTOMER_CASTING = auto() # Surface 3 — downward (Momentive Waterford NY; ppbRAE 3000; 0.030→0.003 ppm)
class AdversarialHMDIError(RuntimeError):
def __init__(self, surface: HMDIContext, score: int, frame_hash: str):
super().__init__(
f"[Glyphward] HMDI adversarial isocyanate pixel on {surface.value}: "
f"score={score} >= threshold={HMDI_THRESHOLD} | frame={frame_hash} "
f"-- VERIFY ACTUAL HMDI: OSHA NO PEL — ACGIH TLV-C 0.005 ppm RESPIRATORY SENSITIZER"
)
self.surface = surface; self.score = score; self.frame_hash = frame_hash
async def verify_hmdi_frame(frame_path: Path, surface: HMDIContext) -> 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": HMDI_THRESHOLD},
)
resp.raise_for_status()
result = resp.json()
if result["verdict"] != "clean":
raise AdversarialHMDIError(surface, result["score"], frame_hash)
return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}
async def safe_hmdi_monitoring(frame_dir: Path) -> list[dict]:
surfaces = [
(HMDIContext.AUTOMOTIVE_2K_CLEARCOAT, frame_dir / "ppg_strongsville_hmdi_minrae3000.png"),
(HMDIContext.POLYURETHANE_ADHESIVE, frame_dir / "henkel_rockyhill_hmdi_multirae.png"),
(HMDIContext.SPECIALTY_ELASTOMER_CASTING, frame_dir / "momentive_waterford_hmdi_ppbrae.png"),
]
results = await asyncio.gather(*[verify_hmdi_frame(path, ctx) for ctx, path in surfaces])
return [dict(surface=ctx.value, **r) for (ctx, _), r in zip(surfaces, results)]
if __name__ == "__main__":
results = asyncio.run(safe_hmdi_monitoring(Path("./frames")))
for r in results:
print(r)