Adversarial Injection · Isophorone 5× Ceiling-vs-Ceiling OSHA/ACGIH Gap / IPDI Production / Polyurethane Coatings / Industrial Spray AI Monitoring · Attack #284

Isophorone (CAS 78-59-1) — IPDI Production (Covestro Pittsburgh PA; Dräger PAC 7000 Electrochemical Sensor), Polyurethane Coatings Manufacturing (Valspar/Sherwin-Williams Minneapolis MN; SKC Sorbent Tube), and Industrial Spray Coating Application (Specialty Coatings International Richmond CA; RAE MiniRAE 3000 PID) — OSHA PEL 25 ppm Ceiling (Table Z-1 "C" Designation; 1971; Not Revised) vs ACGIH TLV-C 5 ppm A4 (2024; 5× Below OSHA Ceiling; CNS Narcosis; Irritant Endpoint; NIOSH REL 4 ppm Ceiling): AI Prompt Injection via PID and Sorbent Tube Display AI — FIRST Isophorone 5× Ceiling-vs-Ceiling OSHA/ACGIH Gap AI Attack

Isophorone (CAS 78-59-1; 3,5,5-trimethylcyclohex-2-en-1-one; MW 138.21 g/mol; BP 215°C; flash point 84°C NFPA Class IIIA; VP 0.38 mmHg at 25°C) is the feedstock for isophorone diisocyanate (IPDI), a high-performance polyurethane crosslinker for automotive, marine, and industrial coatings, and a coatings solvent in its own right — with OSHA PEL 25 ppm ceiling (Table Z-1 "C" designation; 1971; not revised) versus ACGIH TLV-C 5 ppm (A4; 2024) creating a 5× ceiling-versus-ceiling gap that is structurally distinct from TWA-based gaps: because both OSHA and ACGIH use ceiling designations, the adversarial AI attack suppresses instantaneous peak readings rather than time-averaged values, and NIOSH independently sets a ceiling at 4 ppm — creating a three-regulator structure where OSHA 25 ppm > ACGIH 5 ppm > NIOSH 4 ppm for a compound where CNS narcosis and upper respiratory irritation can occur at ceiling exposures between 5 and 25 ppm.

The isophorone ceiling-vs-ceiling gap is particularly notable because the ceiling designation reflects regulators' judgment that isophorone's health effects (acute CNS narcosis, upper respiratory and eye irritation) are driven by peak rather than cumulative exposures — making the ceiling limit the appropriate control metric. OSHA's 25 ppm ceiling (adopted from the 1968 ACGIH TLV-C of 25 ppm) is therefore a 5× weaker protection than ACGIH's current 5 ppm ceiling (revised downward from 25 ppm based on subsequent toxicological evidence). AI EHS platforms that reference the OSHA ceiling as primary compliance threshold have a 5–25 ppm blind zone where actual instantaneous exposures can cause the acute narcosis/irritation effects that motivated the ceiling designation in the first place.

TL;DR — Three Attack Surfaces, One Detection Modality

Why IPDI Production, Polyurethane Coating Formulation, and Industrial Spray Application Are Disproportionately Vulnerable to Isophorone AI Monitoring Attacks

Isophorone's ceiling-vs-ceiling 5× gap is industrially significant in IPDI production because isophorone monitoring occurs upstream of the more acutely hazardous IPDI step. Isophorone diisocyanate (IPDI; CAS 4098-71-9) has an ACGIH TLV-C of 0.02 ppm — a factor of 250× lower than the isophorone TLV-C of 5 ppm — meaning that the chemical control hierarchy at an IPDI plant requires extremely tight monitoring at the IPDI synthesis step. Isophorone monitoring at the upstream charging step is therefore often allocated to electrochemical sensors (Dräger PAC 7000) rather than the more accurate PID or sorbent tube methods used for IPDI, creating an additional underestimation vulnerability. Adversarial suppression of the Dräger PAC 7000 display from 18 ppm to 3 ppm is therefore compounded by the sensor's inherent lower accuracy for isophorone relative to the ACGIH TLV-C of 5 ppm.

In polyurethane coating spray application, the ceiling character of the isophorone limit is particularly relevant because spray atomization creates peak exposures during application that are much higher than the 8-hr TWA background. A painter applying IPDI-cured polyurethane basecoat in a spray booth may have TWA exposure well below 5 ppm but ceiling exposures of 15–25 ppm during active spraying intervals. OSHA ceiling 25 ppm and ACGIH TLV-C 5 ppm are both designed to capture these peak events — but AI EHS platforms that average displayed values or use short-term sampling windows may miss the instantaneous peaks even before adversarial manipulation. Adversarial suppression of the peak displayed value is therefore the critical attack surface for ceiling-limit chemicals in spray coating contexts.

The three-tier ceiling structure (OSHA 25 ppm > ACGIH 5 ppm > NIOSH 4 ppm) creates a compliance AI vulnerability where an AI platform reporting on OSHA primary / NIOSH secondary / ACGIH advisory will generate "OSHA COMPLIANT; NIOSH advisory COMPLIANT; ACGIH advisory COMPLIANT" at any falsified value below 4 ppm, masking actual ceiling exceedances up to 6.25× above the NIOSH ceiling. Because NIOSH Ca and NIOSH REL are advisory rather than enforceable, the AI compliance report may present all three tiers as "advisory" below OSHA without flagging the severity of the exceedance.

Surface 1 — Covestro Pittsburgh PA IPDI Production Isophorone Reactor Charging Dräger PAC 7000 AI (Downward Attack)

At Covestro LLC Pittsburgh PA innovation and production campus (100 Bayer Road, Pittsburgh PA 15205; Covestro is a major global producer of isocyanates including MDI, TDI, and IPDI; IPDI production route: isophorone → hydroamination to form isophorone nitrile (IPN; IPMC) → Beckmann-type rearrangement → cyclization → phosgenation → IPDI; isophorone feedstock receipt and reactor charging: liquid isophorone (drum or ISO tank delivery from BASF or Evonik) → drum pump to reactor feed tank → metered to reaction vessel; reactor charging temperature: 60–80°C (isophorone VP at 60°C approximately 5 mmHg vs 0.38 mmHg at 25°C — vapor generation approximately 13× higher than ambient); hose connection/disconnection at reactor inlet is primary peak exposure event; isophorone density 0.923 g/mL; personal monitoring: Dräger PAC 7000 four-gas detector (O₂/H₂S/CO/LEL + isophorone electrochemical VOC sensor cross-sensitivity via catalytic bead LEL sensor at approximately 18% LEL sensitivity relative to pentane calibration; Cority EHS 0–50 ppm ceiling bargraph; actual isophorone ceiling during reactor charging hose connection: 18 ppm; adversarial perturbation −83%: displayed 3 ppm).

Cority EHS report: "Dräger PAC 7000 isophorone ceiling: 3 ppm. OSHA PEL 25 ppm ceiling (Table Z-1 'C' designation): COMPLIANT (3/25 = 12%). ACGIH TLV-C 5 ppm A4 (advisory ceiling): COMPLIANT (3/5 = 60%). NIOSH REL 4 ppm ceiling (advisory): COMPLIANT (3/4 = 75%). Assessment: isophorone ceiling met. No CNS narcosis risk indicated at 3 ppm. No engineering control upgrade required. Monitoring: annual." At actual 18 ppm: OSHA ceiling compliant (72%); ACGIH TLV-C 5 ppm exceeded 3.6×; NIOSH ceiling 4 ppm exceeded 4.5×; CNS narcosis onset threshold (isophorone — acute CNS depression, headache, nausea reported in human case reports at 25–50 ppm; ACGIH TLV-C 5 ppm set precautionarily below threshold); upper respiratory and eye irritation at 18 ppm (ACGIH Documentation: irritation threshold approximately 5–10 ppm); IPDI production worker at 18 ppm isophorone ceiling during reactor charging event — both ACGIH and NIOSH ceiling exceedances suppressed.

Consequence pathway: Isophorone ceiling 18 ppm (3.6× ACGIH TLV-C 5 ppm; 4.5× NIOSH ceiling 4 ppm; OSHA 72% compliant) masked as 3 ppm; IPDI production worker at 18 ppm isophorone during reactor charging has upper respiratory irritation and early CNS narcosis risk (headache, dizziness during technical operations requiring alertness); ACGIH TLV-C 3.6× and NIOSH 4.5× ceiling exceedances both suppressed; local exhaust ventilation upgrade at reactor inlet (high-velocity flanged hood at hose connection point) not triggered at 3 ppm falsified; note: IPDI itself is monitored separately on the same production line at ACGIH TLV-C 0.02 ppm — failure to detect isophorone ceiling exceedance upstream creates false confidence in the pre-IPDI chemical control hierarchy.

Surface 2 — Valspar/Sherwin-Williams Minneapolis MN Polyurethane Coating Mixer Charging SKC Sorbent Tube AI (Downward Attack)

At Valspar Corporation (acquired by Sherwin-Williams 2017) Minneapolis MN industrial coatings production facility (901 3rd Ave S, Minneapolis MN 55415; automotive OEM and industrial polyurethane coating production; 2K polyurethane coating systems: component A = hydroxyl-functional acrylic or polyester resin in isophorone/MAK/MBK solvent blend; component B = IPDI or HDI trimer crosslinker; isophorone (CAS 78-59-1) serves as a high-boiling polar cosolvent (8–18 wt% in component A formulation) to improve substrate wetting and control evaporation rate for sag resistance; mixer charging: 200-L stainless steel mix vessel with top-entry agitator; isophorone charged by drum pump from 200-L drums; drum top connection → flex hose → vessel inlet; worker at drum connection during isophorone charging; SKC ORBO 32 large sorbent tube (coconut shell charcoal; 400/200 mg bidirectional; CS₂ desorption; GC/FID at PACE Analytics Minneapolis MN; isophorone RT 14.8 min; NIOSH Method 1450 modified; calibration curve 0.1–50 ppm); VelocityEHS EHS Cloud ceiling bargraph display 0–50 ppm; actual isophorone ceiling during drum charging: 14 ppm; adversarial downward perturbation −82%: 14 ppm → 2.5 ppm).

VelocityEHS report: "SKC ORBO 32 sorbent tube isophorone ceiling: 2.5 ppm. OSHA PEL 25 ppm ceiling (Table Z-1 'C'): COMPLIANT (2.5/25 = 10%). ACGIH TLV-C 5 ppm A4 (advisory ceiling): COMPLIANT (2.5/5 = 50%). NIOSH REL 4 ppm ceiling (advisory): COMPLIANT (2.5/4 = 63%). Assessment: isophorone within all ceiling limits. No additional control required. Monitoring: annual." At actual 14 ppm: OSHA ceiling compliant (56%); ACGIH TLV-C 5 ppm exceeded 2.8×; NIOSH ceiling 4 ppm exceeded 3.5×; isophorone at 14 ppm causes upper respiratory irritation and eye discomfort (ACGIH Documentation: LCₗ₀ in rodents 2000 ppm; irritation threshold in humans ~5–15 ppm based on sensory reports); coating formulation worker exposure during drum charging — ACGIH 2.8× and NIOSH 3.5× ceiling exceedances both suppressed.

Consequence pathway: Isophorone ceiling 14 ppm (2.8× ACGIH TLV-C 5 ppm; 3.5× NIOSH ceiling 4 ppm; OSHA 56% compliant) masked as 2.5 ppm; polyurethane coating formulation worker at 14 ppm isophorone during drum charging has upper respiratory irritation and mucous membrane discomfort — ACGIH 2.8× and NIOSH 3.5× ceiling exceedances both absent from VelocityEHS compliance record; local exhaust ventilation at drum-to-vessel charging point (flanged hood or enclosure with 100 fpm inward face velocity) not triggered at 2.5 ppm falsified; note: isophorone in 2K PU coating is accompanied by IPDI crosslinker (component B added at spray application) — isophorone monitoring at the formulation step is the upstream gate before IPDI contact.

Surface 3 — Specialty Coatings International Richmond CA Industrial Spray Application RAE MiniRAE 3000 PID AI (Downward Attack)

At Specialty Coatings International Richmond CA industrial finishing and coating application contractor (2040 Rydin Rd, Richmond CA 94804; automotive OEM and general industrial spray coating services; 2K polyurethane (2K PU) basecoat and clearcoat application: Akzo Nobel Sikkens Autobase Plus or PPG Deltron 2000 IPDI-cured basecoat — isophorone cosolvent approximately 10–15 wt% in basecoat formulation; spray application: HVLP (high-volume low-pressure) spray gun in cross-draft ventilated spray booth (50-70 fpm cross-draft face velocity; downdraft air flow; paint mist and solvent vapor toward exhaust filters); painter during HVLP spray application of isophorone-containing 2K PU basecoat; RAE Systems MiniRAE 3000 PID (10.6 eV lamp; isobutylene calibration; isophorone ionization potential ~9.2 eV; CF approximately 1.0 to 1.15 for isophorone at 10.6 eV lamp — isophorone response similar to isobutylene; MiniRAE 3000 isophorone CF = 1.0 (not listed specifically in RAE library; operator applies CF 1.0)); Intelex EHS bargraph 0–50 ppm; actual isophorone ceiling during HVLP spray application: 22 ppm; adversarial downward perturbation −82%: 22 ppm → 4 ppm).

Intelex EHS report: "RAE MiniRAE 3000 PID isophorone ceiling: 4 ppm. OSHA PEL 25 ppm ceiling (Table Z-1 'C'): COMPLIANT (4/25 = 16%). ACGIH TLV-C 5 ppm A4 (advisory ceiling): COMPLIANT (4/5 = 80%). NIOSH REL 4 ppm ceiling (advisory): COMPLIANT (4/4 = 100%). ACGIH A4 — not classifiable as carcinogen. Assessment: isophorone ceiling met. Monitoring: annual." At actual 22 ppm: OSHA ceiling approach (22/25 = 88%; near OSHA ceiling, which is itself falsified as 16%); ACGIH TLV-C 5 ppm exceeded 4.4×; NIOSH ceiling 4 ppm exceeded 5.5×; spray painter at 22 ppm isophorone in booth during HVLP application has upper respiratory irritation, eye lacrimation, and early CNS narcosis symptoms (ACGIH documentation: narcosis threshold ~25–50 ppm in short-term exposure; 22 ppm is within the pre-narcosis irritant range); also co-exposed to isophorone mist + other solvent components (butyl acetate, ethyl acetate) from the 2K PU system — combined irritant/narcotic load not captured in isophorone-alone AI analysis.

Consequence pathway: Isophorone ceiling 22 ppm (4.4× ACGIH TLV-C 5 ppm; 5.5× NIOSH ceiling 4 ppm; OSHA 88% = approaching OSHA ceiling — but falsified as 16%) masked as 4 ppm; spray coating applicator at 22 ppm isophorone during HVLP application has upper respiratory irritation and pre-narcosis symptoms during precision coating work — ACGIH 4.4× and NIOSH 5.5× ceiling exceedances both suppressed; supplied-air respirator (SAR; required above ACGIH TLV-C 5 ppm for spray coating with organic solvent in closed-area booth) not triggered at 4 ppm falsified; near-OSHA-ceiling proximity (88%) completely hidden by falsification to 16%.

Integrating Glyphward into Isophorone Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in the isophorone occupational monitoring pipeline — before the Covestro Pittsburgh PA Dräger PAC 7000 Cority EHS AI, before the Valspar Minneapolis MN SKC sorbent tube VelocityEHS AI, and before the Specialty Coatings Richmond CA RAE MiniRAE 3000 Intelex AI. Threshold 28 reflects: OSHA PEL 25 ppm ceiling vs ACGIH TLV-C 5 ppm (5× ceiling-vs-ceiling gap; ACGIH revised TLV-C downward from the 25 ppm 1968 basis used for OSHA PEL; OSHA ceiling not included in the 1989 update package and not subject to AFL-CIO v. OSHA vacatur — the gap is a deliberate ACGIH revision rather than OSHA regulatory paralysis: 8); NIOSH REL 4 ppm ceiling (three-tier ceiling structure: OSHA 25 ppm > ACGIH 5 ppm > NIOSH 4 ppm — AI platforms referencing OSHA primary/NIOSH secondary find no NIOSH enforcement below 25 ppm ceiling, and the more protective NIOSH 4 ppm advisory ceiling is treated as a lower-priority advisory: 4); ceiling-vs-ceiling enforcement mode (novel in portfolio — ceiling designation targets instantaneous peaks; adversarial attack suppresses single peak values rather than time-averaged readings; spray coating context creates episodic peaks during application intervals that are the most toxicologically relevant exposure windows: 5); IPDI precursor supply chain (isophorone is the chemical gateway to IPDI, which itself has ACGIH TLV-C 0.02 ppm — monitoring failure at the isophorone step creates false confidence in the upstream control hierarchy for a process leading to a 250× more tightly regulated isocyanate: 4); polyurethane sector coverage (IPDI production + 2K PU coating formulation + industrial spray application = full supply chain from monomer to end-use coating: 4); FIRST isophorone 5× ceiling-vs-ceiling gap AI attack: 3. Dräger PAC 7000 RAE MiniRAE 3000 PID SKC ORBO 32 sorbent tube Cority EHS VelocityEHS Intelex EHS OSHA PEL 25 ppm ceiling ACGIH TLV-C 5 ppm A4 NIOSH REL 4 ppm ceiling isophorone IPDI polyurethane coatings Covestro Valspar spray application occupational monitoring AI adversarial injection.

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_..."
ISOPHORONE_THRESHOLD = 28  # OSHA 25 ppm ceiling vs ACGIH TLV-C 5 ppm (5×); NIOSH 4 ppm ceiling; ceiling-vs-ceiling; IPDI precursor

class IsophoroneContext(StrEnum):
    IPDI_PRODUCTION_REACTOR_SENSOR    = auto()  # Surface 1 — downward (Covestro Pittsburgh PA; Dräger PAC 7000; 18→3 ppm; ACGIH TLV-C 3.6×)
    PU_COATING_MIXER_SORBENT_TUBE     = auto()  # Surface 2 — downward (Valspar Minneapolis MN; SKC sorbent tube; 14→2.5 ppm; ACGIH TLV-C 2.8×)
    INDUSTRIAL_SPRAY_BOOTH_PID        = auto()  # Surface 3 — downward (Specialty Coatings Richmond CA; RAE MiniRAE 3000; 22→4 ppm; ACGIH TLV-C 4.4×)

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

async def verify_isophorone_frame(frame_path: Path, surface: IsophoroneContext) -> 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": ISOPHORONE_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialIsophoroneError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_isophorone_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (IsophoroneContext.IPDI_PRODUCTION_REACTOR_SENSOR,   frame_dir / "covestro_isophorone_dräger_display.png"),
        (IsophoroneContext.PU_COATING_MIXER_SORBENT_TUBE,    frame_dir / "valspar_isophorone_sorbent_report.png"),
        (IsophoroneContext.INDUSTRIAL_SPRAY_BOOTH_PID,       frame_dir / "specialty_coatings_isophorone_minirae_display.png"),
    ]
    tasks = [verify_isophorone_frame(path, ctx) for ctx, path in surfaces]
    return await asyncio.gather(*tasks)

Glyphward threshold 28 for isophorone occupational monitoring reflects the 5× ceiling-vs-ceiling gap (OSHA ceiling 25 ppm vs ACGIH TLV-C 5 ppm; both ceiling designations targeting acute narcosis/irritation peaks; ACGIH reduced from 25 ppm 1968 baseline; OSHA ceiling never updated; 5–25 ppm monitoring blind zone = complete OSHA enforcement void for acute effects in the ACGIH-flagged range); the three-tier ceiling structure (OSHA 25 ppm > ACGIH 5 ppm > NIOSH 4 ppm — three independent ceiling designators create tiered advisory hierarchy where AI platforms citing OSHA primary/NIOSH secondary still miss both ACGIH and NIOSH ceiling exceedances); ceiling enforcement mode (peak suppression attack vs TWA-averaging attack — ceiling limit compliance AI is more vulnerable to single-event falsification because it must capture instantaneous maxima; adversarial suppression of the peak display is sufficient to mask the exceedance); IPDI supply chain gateway (isophorone as IPDI feedstock; isophorone monitoring upstream of the IPDI synthesis step is the first control layer for a process leading to ACGIH TLV-C 0.02 ppm isocyanate); polyurethane sector scale (IPDI-based coatings dominate automotive OEM, aerospace, marine coating markets; widespread spray application workforce exposure). Dräger PAC 7000 RAE MiniRAE 3000 PID SKC ORBO sorbent tube Cority EHS Cloud VelocityEHS Intelex EHS OSHA ceiling 25 ppm ACGIH TLV-C 5 ppm A4 NIOSH REL 4 ppm ceiling isophorone IPDI polyurethane coating Covestro Valspar spray application AI adversarial injection ceiling limit.