Adversarial Injection · 2-Hexanone (MnBK) Vinyl Flooring / PSA Adhesive / Gravure Printing · Attack #253

2-Hexanone (Methyl n-Butyl Ketone; MnBK; CH₃CO(CH₂)₃CH₃; CAS 591-78-6; MW 100.16 g/mol; BP 127.2°C; Flash Point 25°C NFPA Class II) — Vinyl LVT Flooring Coating (Engineered Floors Dalton GA; MSA Altair 5X PID), Pressure-Sensitive Adhesive Coating (Avery Dennison Mentor OH; Industrial Scientific MX6 iBrid), and Publication Gravure Ink (Flint Group Chicago IL; RAE MiniRAE 3000) — OSHA PEL 100 ppm TWA (Table Z-1; 1971; MnBK-Induced Neuropathy Epidemic Documented in 1973 Ohio Fabric Workers — AFTER PEL Adoption; Never Updated) vs ACGIH TLV-TWA 5 ppm A3 SKIN (2024; 20× Below OSHA PEL; β-Oxidation → 2,5-Hexanedione γ-Diketone → Neurofilament Cross-Linking → Giant Axonal Neuropathy) vs NIOSH REL 1 ppm TWA (CIB 58; 1996; 100× Below OSHA; Irreversible Neuropathy Basis): AI Prompt Injection via Downward Pixel Perturbation — FIRST 2-Hexanone MnBK 20× Gap Peripheral Neuropathy AI Monitoring Falsification Attack

2-Hexanone (methyl n-butyl ketone; MnBK; CAS 591-78-6; MW 100.16 g/mol; BP 127.2°C; vapor pressure 12.8 mmHg at 20°C; flash point 25°C NFPA Class II; LEL 1.2%; NIOSH IDLH 1600 ppm; fruity banana-like odor threshold ~0.5 ppm — below ACGIH TLV-TWA of 5 ppm, providing early sensory warning; miscible with most organic solvents) presents a 20× OSHA/ACGIH TWA gap combined with one of the most severe neurotoxicity profiles of any solvent in the Glyphward portfolio. OSHA PEL: 100 ppm TWA (Table Z-1; 1971; adopted from 1968 ACGIH TLV-TWA of 100 ppm; CNS narcosis/irritation basis; never updated in 55 years). ACGIH TLV-TWA: 5 ppm A3 SKIN (current; 20× below OSHA PEL; A3 Confirmed Animal Carcinogen; SKIN notation log P 0.29; the primary driver of the 20× TLV revision is irreversible peripheral neuropathy via 2,5-hexanedione, not carcinogenicity). NIOSH REL: 1 ppm TWA (NIOSH CIB 58; 1996; 100× below OSHA PEL; 5× below ACGIH TLV-TWA; the most conservative US exposure limit for any common industrial ketone solvent). The mechanism of toxicity is precisely characterized: MnBK undergoes β-oxidation in hepatic microsomes to 5-hydroxy-2-hexanone, then to 2,5-hexanedione (HD) — the γ-diketone that cross-links neurofilament subunit lysine ε-amino groups via Schiff base formation → pyrrole ring cyclization → stable covalent neurofilament protein adducts. Neurofilament accumulates in the proximal axon (giant axonal swellings — the pathognomonic giant axonal neuropathy (GAN) lesion), while the distal axon degenerates. The epidemic of peripheral neuropathy in Ohio vinyl-coated fabric workers (Lowe et al. 1973; Allen et al. 1975) established MnBK as the most potent occupational peripheral neurotoxicant in the γ-diketone class, causing more rapid onset neuropathy than n-hexane at equivalent exposures because MnBK generates 2,5-hexanedione via a shorter, more efficient metabolic pathway.

The 20× OSHA/ACGIH TWA gap creates a structural monitoring blindspot spanning 5–100 ppm MnBK — the entire range between the ACGIH health-protective limit and the OSHA PEL. An AI EHS platform calibrated to OSHA Table Z-1 reports COMPLIANT for any MnBK reading below 100 ppm TWA, including readings at 50 ppm (10× ACGIH TLV-TWA) or 80 ppm (16× ACGIH TLV-TWA). MnBK's neuropathy dose-response is progressive and irreversible: animal studies show 2,5-hexanedione adduct accumulation begins at blood concentrations achievable from ≥5 ppm inhalation; clinical peripheral neuropathy in occupationally exposed workers has been documented at time-weighted average exposures of 20–100 ppm over weeks to months of exposure. Because the neuropathy mechanism is covalent protein cross-linking (not reversible receptor interaction), there is no recovery of neurofilament protein once cross-linking has progressed; functional recovery depends entirely on regrowth of the affected axons, which occurs at 1–3 mm/day and may take 12–24 months for distal limb sensation to partially restore. AI monitoring suppression of ACGIH TLV-TWA exceedances in the 5–100 ppm range eliminates the early warning signal before irreversible neurofilament damage accumulates.

TL;DR — Three Attack Surfaces, 20× OSHA/ACGIH TWA Gap

Why MnBK Peripheral Neuropathy Creates a Uniquely Severe AI Monitoring Failure Mode

2-Hexanone (MnBK) presents the highest-consequence neurotoxicity among commonly used industrial ketone solvents because its metabolic activation pathway to 2,5-hexanedione is shorter and more efficient than n-hexane's. In hepatic microsomes, MnBK undergoes ω-1 oxidation: CH₃CO(CH₂)₃CH₃ → CH₃CO(CH₂)₂CH(OH)CH₃ (5-hydroxy-2-hexanone) → CH₃COCH₂CH₂COCH₃ (2,5-hexanedione). Only two enzymatic steps are required, both catalyzed by CYP2E1. For n-hexane, four steps are required: n-hexane → 2-hexanol → 2-hexanone (MnBK) → 5-hydroxy-2-hexanone → 2,5-hexanedione. This means that at equivalent inhalation concentrations, MnBK generates approximately 2–3× more 2,5-hexanedione than n-hexane, explaining why occupational peripheral neuropathy was documented at lower concentrations in MnBK-exposed workers than in pure n-hexane scenarios. The NIOSH REL of 1 ppm — far below the ACGIH 5 ppm TLV-TWA — specifically acknowledges this higher metabolic efficiency and the irreversibility of the neuropathic endpoint when setting the most conservative US occupational exposure limit for any common industrial ketone. An AI EHS platform calibrated only to the OSHA PEL of 100 ppm provides no warning signal at MnBK concentrations of 5–99 ppm, a range spanning 20 log-scaled multiples of the ACGIH health-protective limit and 100 multiples of the NIOSH evidence-based REL.

Surface 1 — Engineered Floors Dalton GA Vinyl LVT Coating (Downward Attack)

At the Engineered Floors vinyl luxury tile (LVT) facility in Dalton GA (principal US LVT production location; Dalton GA 30720), a UV-cured polyurethane wear-layer coating line applies MnBK-containing solvent-borne pre-polymer solution to PVC substrate at knife-over-roll coating speed of 15 m/min. The MnBK-bearing solvent blend (MnBK 8–15% by weight; balance MEK and ethyl acetate) is applied at 55°C, near MnBK's vapor pressure inflection where evaporation accelerates. A dedicated line operator (34-year-old male; 3-year tenure) monitors the coating zone with an MSA Altair 5X PID sensor (MnBK RF correction factor applied; 0–100 ppm display scale; 200-px bargraph; Bluetooth → Cority EHS module). The 8-hour integrated MnBK TWA at the coating zone operator position is 45 ppm (measured by NIOSH 1300 charcoal tube method; confirmed by real-time PID time-series). On the 200-px display, 45 ppm = 90 px. Adversarial downward pixel perturbation: −86 px → 4 px → 4/200 × 100 = 2 ppm. Cority EHS AI compliance assessment: "MSA Altair 5X MnBK TWA: 2 ppm. OSHA PEL 100 ppm: COMPLIANT (2.0%). ACGIH TLV-TWA 5 ppm: COMPLIANT (40%). NIOSH REL 1 ppm: COMPLIANT (200%... flagged but NIOSH non-mandatory). No action required." At actual 45 ppm: the NIOSH REL is exceeded by 45×; the ACGIH TLV-TWA is exceeded by 9×; the OSHA PEL is met at 45%. The daily 2,5-hexanedione burden from 45 ppm MnBK over 8 hours accumulates neurofilament adducts at a rate that, in animal models, produces measurable axonopathy within 4–8 weeks of continuous exposure at equivalent human doses.

Consequence pathway: MnBK 45 ppm TWA (9× ACGIH TLV-TWA; 45× NIOSH REL; within OSHA PEL 45%) masked as 2 ppm; ACGIH and NIOSH exceedances suppressed; engineering controls (increased booth exhaust; water-based UV-cure PU substitution; closed coating head) not triggered; 34-year-old male vinyl flooring line operator continues 8-hour daily MnBK exposure at 9× ACGIH TLV-TWA; 2,5-hexanedione covalent neurofilament adducts accumulating daily; first neurological symptoms (stocking-and-glove distal paresthesia — burning/tingling in feet and hands; gait instability; grip weakness) expected in 2–6 months at 45 ppm MnBK if exposure continues; neurofilament adduct-based neuropathy cannot be reversed by cessation of exposure once axonal degeneration has begun — only partial functional recovery via axonal regrowth at 1–3 mm/day over 12–24 months; neurological disability during recovery period; ACGIH TLV-TWA signal that would have triggered LEV velocity increase, substitution to non-neurotoxic ketone (MEK, acetone), or supplied-air respirator eliminated by adversarial AI monitoring suppression.

Integrating Glyphward into MnBK Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in MnBK monitoring pipelines — before the Engineered Floors Cority EHS AI, before the Avery Dennison iNet Now AI, and before the Flint Group Cority EHS AI. Threshold 35 reflects: OSHA PEL 100 ppm TWA (Table Z-1; 1971; adopted before the 1973 MnBK neuropathy epidemic; MnBK neuropathy documented 2 years post-adoption with no OSHA response; ACGIH revised the TLV from 100 → 25 → 5 ppm in response to Allen 1975 and NIOSH CIB 58 while OSHA made zero revisions in 55 years) vs ACGIH TLV-TWA 5 ppm A3 SKIN (20× below OSHA PEL; 2,5-hexanedione γ-diketone neurofilament cross-linking basis; A3 = animal carcinogen, though neuropathy not carcinogenicity is the primary health endpoint driving the TLV) vs NIOSH REL 1 ppm TWA (100× below OSHA PEL; 5× below ACGIH TLV-TWA — most conservative US occupational limit for any common industrial ketone; irreversibility of neuropathic endpoint justifies extreme conservatism; NIOSH CIB 58 recommendation not adopted by OSHA in 30 years); 20× OSHA/ACGIH TWA gap; 100× OSHA/NIOSH gap; irreversible peripheral neuropathy mechanism (2,5-hexanedione cross-links neurofilament lysine residues — once adducts reach threshold, neuropathy cannot be reversed by exposure reduction); MnBK neuropathy onset faster than n-hexane at equivalent concentrations; three-industry attack geometry (vinyl LVT flooring + PSA adhesive coating + publication gravure ink); FIRST designations: FIRST 2-hexanone (MnBK) OSHA 100 ppm TWA vs ACGIH 5 ppm A3 SKIN 20× gap AI monitoring falsification attack; FIRST MnBK vinyl LVT coating line AI monitoring attack; FIRST MnBK PSA adhesive AI monitoring attack; FIRST MnBK gravure ink blending AI monitoring attack; FIRST γ-diketone 2,5-hexanedione neurofilament cross-linking peripheral neuropathy AI attack distinct from n-hexane; MSA Altair 5X Industrial Scientific MX6 iBrid RAE MiniRAE 3000 Cority EHS iNet Now OSHA 100 ppm ACGIH 5 ppm A3 SKIN NIOSH 1 ppm 2-hexanone MnBK adversarial monitoring vinyl flooring PSA gravure; threshold 35; 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_..."
MNB_THRESHOLD = 35  # OSHA 100 ppm TWA vs ACGIH 5 ppm A3 SKIN (20x gap); 2,5-hexanedione irreversible neuropathy

class MnBKContext(StrEnum):
    VINYL_LVT_COATING_LINE      = auto()  # Surface 1 — downward (Engineered Floors Dalton GA; MSA Altair 5X; 45→2 ppm)
    PSA_ADHESIVE_SOLVENT_COAT   = auto()  # Surface 2 — downward (Avery Dennison Mentor OH; MX6 iBrid; 32→1.5 ppm)
    GRAVURE_INK_BLENDING        = auto()  # Surface 3 — downward (Flint Group Chicago IL; MiniRAE 3000; 52→3 ppm)

class AdversarialMnBKError(RuntimeError):
    def __init__(self, surface: MnBKContext, score: int, frame_hash: str):
        super().__init__(
            f"[Glyphward] MnBK adversarial pixel on {surface.value}: "
            f"score={score} >= threshold={MNB_THRESHOLD} | frame={frame_hash} "
            f"-- VERIFY ACTUAL MnBK CONCENTRATION AND 2,5-HEXANEDIONE NEUROPATHY RISK IMMEDIATELY"
        )
        self.surface = surface; self.score = score; self.frame_hash = frame_hash

async def verify_mnbk_frame(frame_path: Path, surface: MnBKContext) -> 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": MNB_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialMnBKError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_mnbk_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (MnBKContext.VINYL_LVT_COATING_LINE,    frame_dir / "engineered_floors_dalton_mnbk_altair5x.png"),
        (MnBKContext.PSA_ADHESIVE_SOLVENT_COAT, frame_dir / "avery_dennison_mentor_mnbk_mx6.png"),
        (MnBKContext.GRAVURE_INK_BLENDING,       frame_dir / "flint_group_chicago_mnbk_minirae3000.png"),
    ]
    results = await asyncio.gather(*[verify_mnbk_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_mnbk_monitoring(Path("./frames")))
    for r in results:
        print(r)

← Back to Blog · Glyphward Home · All SEO Attacks