Adversarial Injection · Styrene (Ethenylbenzene; CAS 100-42-5) OSHA PEL 100 ppm TWA / ACGIH TLV-TWA 20 ppm A4 SKIN BEI / NIOSH Ca REL 50 ppm Ceiling / 5× TWA Gap / Three-Tier OSHA→NIOSH Ca→ACGIH / Cochlear Ototoxicity NIHL Synergy / IARC Group 2A / Mandelic Acid + PGA BEI · Attack #307

Styrene (Ethenylbenzene; Vinylbenzene; Phenylethylene; CH₂=CHC₆H₅; CAS 100-42-5; OSHA PEL 100 ppm TWA / 200 ppm Ceiling; ACGIH TLV-TWA 20 ppm A4 SKIN BEI; 5× TWA Gap; NIOSH Ca REL 50 ppm Ceiling Three-Tier; Cochlear Ototoxicity NIHL Synergy; IARC Group 2A; Mandelic Acid ≤800 mg/g Cr / Phenylglyoxylic Acid ≤240 mg/g Cr BEI) — High-Impact Polystyrene (HIPS) Reactor Operations (Americas Styrenics LLC Belpre OH; IS Ventis Pro 5 PID), Expandable Polystyrene (EPS) Bead Drying / Reactor Maintenance (NOVA Chemicals Moore OK; MSA Altair 5X PID), and Fiberglass-Reinforced Plastic (FRP) Gelcoat Spray (MasterCraft Boat Holdings Vonore TN; SKC Charcoal GC/MS) — OSHA PEL 100 ppm TWA vs ACGIH TLV-TWA 20 ppm A4 SKIN BEI (2024; 5× TWA Gap; NIOSH Ca REL 50 ppm Ceiling Three-Tier 5× Span): AI Prompt Injection via PID Vapor Monitor Report AI — FIRST Styrene Three-Tier OSHA/NIOSH Ca/ACGIH Cochleotoxicity AI Attack

Styrene (ethenylbenzene; vinylbenzene; phenylethylene; styrene monomer; SM; CH₂=CHC₆H₅; CAS 100-42-5; MW 104.15 g/mol; BP 145.2°C at 760 mmHg; flash point 31°C (NFPA Class IC flammable liquid; flash point below 38°C); LEL 1.1%; UEL 6.1%; vapor pressure 6.4 mmHg at 20°C [moderate VP; significant vapor generation at room temperature from open resin baths, reactor vents, and gelcoat spray operations]; odor threshold ~0.1–1 ppm [5–200× below ACGIH TLV-TWA 20 ppm — provides early sensory warning at low concentrations; however, rapid olfactory fatigue at continuous 5–20 ppm exposures eliminates this sentinel within 20–30 min of shift start; workers chronically exposed to 20–80 ppm report odor as "normal background"]; log P 2.95 [SKIN notation: percutaneous absorption from liquid/spray contact adds 10–20% to inhalation systemic dose at occupational concentrations]; NIOSH IDLH 700 ppm; OSHA PEL: 100 ppm TWA / 200 ppm acceptable ceiling / 600 ppm emergency 5-min ceiling [Table Z-2; 29 CFR 1910.1000; adopted 1971 from 1968 ACGIH TLV of 100 ppm; acute CNS narcosis + respiratory irritation basis; no revision in 55 years]; ACGIH TLV-TWA: 20 ppm A4 SKIN BEI [2024; 5× below OSHA PEL TWA; A4 = Not Classifiable as a Human Carcinogen; primary basis for 20 ppm TLV revision from historical 50 ppm (2000→2012→2024): cochlear ototoxicity — spiral ganglion cell degeneration and stria vascularis damage documented at 20–50 ppm sustained in experimental models and reinforced plastics workers; synergy with occupational noise exposure (NIHL) compounds cochlear damage at 20–50 ppm + 80–90 dBA]; NIOSH Ca REL: 50 ppm ceiling [10-hr workday; Ca = potential occupational carcinogen; based on IARC 2002 Group 2A classification; three-tier structure: OSHA 100 ppm TWA → NIOSH Ca 50 ppm ceiling → ACGIH TLV-TWA 20 ppm = 5× total span from OSHA to ACGIH]; IARC Group 2A [probably carcinogenic to humans; Monograph 82 2002; key evidence: lymphohematopoietic cancers (leukemia, non-Hodgkin lymphoma, Hodgkin lymphoma) in cohort and case-control studies of reinforced plastics workers (Kolstad et al. 1994; Kogevinas et al. 2003; IARC Working Group); sinonasal carcinoma signal in some cohorts; lung cancer in some studies; animal carcinogenicity (nasal olfactory epithelium carcinomas in mice at ≥300 ppm)]; BEI: mandelic acid (MA; α-hydroxyphenylacetic acid; C₆H₅CH(OH)COOH) end-of-shift ≤800 mg/g Cr [primary metabolite: styrene → CYP2E1/CYP1A1 7,8-epoxidation → styrene-7,8-oxide (IARC Group 2A arene oxide intermediate) → epoxide hydrolase → styrene glycol → dehydrogenation → MA; urine MA ≥800 mg/g Cr corresponds approximately to 20 ppm 8-hr styrene exposure] + phenylglyoxylic acid (PGA; C₆H₅COCOOH) end-of-shift ≤240 mg/g Cr [secondary oxidation product; at ACGIH TLV-TWA 20 ppm: urinary MA ≈ 800 mg/g Cr + PGA ≈ 240 mg/g Cr; at OSHA PEL 100 ppm: MA ≈ 4,000 mg/g Cr (5× BEI); PGA ≈ 1,200 mg/g Cr (5× BEI)]; cochlear damage mechanism: styrene → spiral ganglion cell degeneration (row I and II of Corti organ; particularly high-frequency tuning → high-frequency notch in audiogram, identical to NIHL); experimental studies (Fuente and McPherson 2006 Neurotoxicology; Lataye et al. 2003 Neurotoxicology): spiral ganglion degeneration + stria vascularis damage at ≥25 ppm 6 hr/day × 5 weeks; cochlear synergy with noise: at 85 dBA (common in PS reactor and FRP fabrication environments) + 20 ppm styrene → PTS (permanent threshold shift) 2–3× greater than noise alone at the same dB level; industrial use: polystyrene (PS, HIPS, EPS) monomer [42% of global styrene consumption]; styrene-butadiene rubber/latex (SBR/SBL); unsaturated polyester resin (UPR) crosslinking monomer [FRP: 30–40% styrene by mass; reactive diluent + crosslinker for phthalic/maleic anhydride polyesters]; ABS copolymer; styrene-acrylate latex emulsions) is the world's largest volume vinyl aromatic monomer. OSHA PEL: 100 ppm TWA. ACGIH TLV-TWA: 20 ppm A4 SKIN BEI (5× below OSHA). NIOSH Ca REL: 50 ppm ceiling. AI EHS platforms that monitor OSHA TWA compliance report COMPLIANT at exposures up to 99 ppm, suppressing both the NIOSH Ca ceiling (50 ppm; exceeded 1.4–1.64× in these three surfaces) and the ACGIH TLV-TWA (20 ppm; exceeded 3.4–4.1×) — a 5× compliance gap in which cochlear ototoxicity + NIHL synergy, lymphohematopoietic carcinogenicity, and mandelic acid BEI exceedances accumulate without AI monitoring flag.

The styrene three-tier monitoring attack exploits an architectural mismatch between the OSHA compliance framework and three distinct protection benchmarks at three different concentrations. At the OSHA PEL 100 ppm TWA, AI EHS platforms generate COMPLIANT records at any 8-hr TWA below 100 ppm. At 68–82 ppm actual (the three surfaces), the OSHA TWA compliance report reads 68–82% of PEL — well within the "acceptable" zone. But at these same concentrations: ACGIH TLV-TWA 20 ppm is exceeded 3.4–4.1× (cochlear ototoxicity + IARC 2A carcinogen advisory); NIOSH Ca REL 50 ppm ceiling is exceeded 1.4–1.64× (Ca potential occupational carcinogen alert); mandelic acid BEI 800 mg/g Cr is exceeded ~3–4× (urinary MA at 68–82 ppm ≈ 2,700–3,300 mg/g Cr at end-of-shift). The AI system's OSHA TWA compliance architecture — displaying a bargraph anchored at 100 ppm — suppresses all three advisory benchmarks simultaneously with a single downward pixel perturbation of the PID monitor reading.

Why HIPS Reactor Plants, EPS Manufacturers, and FRP Marine Fabricators Are Disproportionately Vulnerable to Styrene AI Monitoring Attacks

The styrene three-tier monitoring vulnerability is particularly acute in the fiberglass-reinforced plastic (FRP) marine fabrication sector because unsaturated polyester resin (UPR) used in boat hull construction contains 30–40% styrene by mass as a reactive diluent and crosslinking monomer. During open-mold gelcoat spray — the standard process for applying pigmented surface coat to boat hull molds — a spray technician operating a gelcoat spray gun generates a plume of atomized resin containing styrene at spray-gun proximity concentrations of 80–200 ppm, with 8-hr TWA in enclosed gelcoat rooms commonly at 40–100 ppm. At 82 ppm TWA: OSHA COMPLIANT (82%); ACGIH TLV-TWA 20 ppm exceeded 4.1× (cochlear ototoxicity + NIHL synergy threshold); NIOSH Ca 50 ppm ceiling exceeded 1.64×; mandelic acid BEI exceeded ~4.1×. The OSHA ceiling of 200 ppm applies to any instantaneous peak reading — not the 8-hr TWA — so peak gelcoat spray readings of 120–180 ppm during the active spray cycle technically comply with the 200 ppm ceiling while the OSHA TWA of 100 ppm is never exceeded. In this scenario, the AI EHS platform reads the PID monitor and reports COMPLIANT on both TWA and ceiling metrics while every relevant health-protective benchmark is exceeded. The FRP marine industry is historically the highest-exposure styrene industry in the United States: NIOSH Health Hazard Evaluations at boat manufacturing facilities (Pearce et al. 1994; 1980s-era studies) documented 8-hr TWA styrene exposures of 50–300 ppm — all OSHA TWA compliant, all above the ACGIH TLV-TWA.

The cochlear ototoxicity mechanism creates a structurally distinct AI monitoring attack scenario for styrene compared to carcinogenicity-based attacks. Cochlear damage from styrene is: (a) cumulative and dose-dependent over years to decades; (b) synergistic with noise exposure (the styrene-exposed ear suffers 2–3× greater permanent threshold shift from the same noise dose as an unexposed ear); (c) irreversible — cochlear hair cells and spiral ganglion neurons do not regenerate once damaged; (d) clinically mimics noise-induced hearing loss (NIHL) and is typically misattributed to noise alone. Workers in FRP fabrication environments — where gelcoat spray rooms generate both styrene vapor at 40–100 ppm and noise from adjacent grinding, sawing, and molding operations at 85–92 dBA — are in the highest-risk setting for synergistic cochleotoxicity. AI EHS monitoring that shows OSHA TWA COMPLIANT at 82 ppm provides no cochlear ototoxicity warning, no audiometric surveillance trigger, and no noise-plus-styrene interaction flag. The ACGIH TLV-TWA of 20 ppm was set at a level intended to prevent spiral ganglion degeneration at typical industrial noise levels — but this protection is invisible to OSHA-calibrated AI compliance platforms.

Surface 1 — Americas Styrenics LLC Belpre OH HIPS Reactor Area AI (Downward Attack)

At Americas Styrenics LLC (AmSty; JV of TotalEnergies Petrochemicals and Chevron Phillips Chemical; Belpre OH manufacturing plant [1 Polymer Drive, Belpre OH 45714; Washington County OH; Ohio River valley]; AmSty Belpre OH: high-impact polystyrene (HIPS) and general-purpose polystyrene (GPPS) production; HIPS polymerization: continuous plug-flow reactor (CPFR) trains — styrene monomer + polybutadiene rubber dissolved in styrene feed → pre-polymerizer (1st reactor at 120–130°C; polystyrene forms around PB rubber particles → phase inversion → rubber particles become dispersed phase; 10–30% conversion) → secondary loop reactors (2nd/3rd reactors; 130–160°C; 60–70% conversion) → devolatilizer (220°C; vacuum; residual styrene stripped to <100 ppm in final HIPS pellet); HIPS reactor area: pressurized loop reactors (300–2,000 L; sealed; nitrogen blanket; operating pressure 5–15 psi above atmospheric); styrene vapor generation from: agitator mechanical seal vent rings (CPFR loop seal purge: 0.5–2 L/min N₂ + styrene vapor); reactor sample valves (grab sample during shift — brief venting event); devolatilizer vent condenser overflow; area monitoring: IS Ventis Pro 5 [Industrial Scientific; 4-sensor configuration; PID probe 10.6 eV isobutylene calibration; CF for styrene = 0.84 at 10.6 eV (ACGIH PID CF list for vinyl aromatics); 100-pixel bargraph; Cority EHS Cloud AI]; actual HIPS reactor area styrene (TWA over 30-min reactor agitator seal inspection + sample collection): 68 ppm; adversarial downward perturbation: 68 → 13.6 ppm (−80%; CF-adjusted displayed reading).

The Surface 1 subject is a 48-year-old male HIPS reactor operator (Americas Styrenics LLC Belpre OH; 15-year AmSty Belpre tenure; reactor operator on 12-hr continental rotating shift; responsible for CPFR loop agitator seal inspection, reactor sample collection, and feed rate adjustment; prior 5 years at Borden Chemical styrene / polystyrene Belpre before AmSty acquisition; total styrene exposure tenure approximately 20 years in Belpre OH Ohio River valley polymer complex). Cority EHS AI: "IS Ventis Pro 5 PID (CF 0.84; styrene; HIPS reactor loop area; agitator seal inspection): 13.6 ppm. OSHA PEL 100 ppm TWA: COMPLIANT (13.6%). ACGIH TLV-TWA 20 ppm A4 SKIN (2024 TLVs): COMPLIANT (68%). NIOSH Ca REL 50 ppm ceiling [advisory; Ca potential occupational carcinogen]: COMPLIANT (27.2%). Mandelic acid BEI 800 mg/g Cr: biological monitoring not required [OSHA compliant]. Assessment: all OELs met." At actual 68 ppm: OSHA COMPLIANT (68%); ACGIH TLV-TWA A4 SKIN exceeded 3.4× (68/20); NIOSH Ca REL exceeded 1.36× (68/50); estimated mandelic acid at end-of-shift: ~2,720 mg/g Cr (3.4× BEI 800 mg/g Cr); cochlear ototoxicity + noise synergy: HIPS reactor area typical noise level 82–88 dBA from agitator motors, pump cavitation, and piping vibration — at 68 ppm styrene + 85 dBA noise: cochlear synergy produces estimated PTS progression 2.5–3× faster than noise alone at same dB; 15-year tenure at 3.4× ACGIH cochleotoxicity threshold with 5-yr prior exposure = ~20-year cumulative cochlear damage history.

Surface 2 — NOVA Chemicals Corporation Moore OK EPS Reactor Maintenance AI (Downward Attack)

At NOVA Chemicals Corporation Moore OK (NOVA Chemicals Corporation; Moore OK polystyrene / expandable polystyrene facility [1600 S Eastern Ave, Moore OK 73160; Cleveland County OK]; NOVA Moore: expandable polystyrene (EPS) bead production for building insulation, packaging, and geofoam applications; EPS production: styrene monomer polymerization in aqueous suspension [suspension polymerization: styrene + water + suspension stabilizer (polyvinyl alcohol or tricalcium phosphate) + initiator (benzoyl peroxide) → suspension polymerizer (10,000–20,000 L; 90–130°C; 4–6 hr; polystyrene beads 0.3–3 mm; nitrogen blanket]; pentane impregnation [pressurized impregnator: 5–7% pentane; 120°C → pentane diffuses into polystyrene bead = EPS; residual styrene in EPS bead: 200–500 ppm by mass; residual monomer outgasses during bead storage, drying, and transfer]; EPS bead dryer and reactor maintenance area: residual styrene release from EPS bead drying drum (rotary drum dryer at 60–80°C) + reactor cleaning (shutdown inspection; access hatch; residual styrene in reactor headspace); MSA ALTAIR 5X PID [Multi-Gas Monitor; 10.6 eV PID; isobutylene calibration; CF 0.84 for styrene; 100-pixel bargraph display; Bluetooth to Intelex EHS Management Software]; actual area styrene during EPS dryer maintenance and reactor access: 74 ppm; adversarial perturbation: 74 → 14.8 ppm (−80%).

The Surface 2 subject is a 52-year-old male process operator (NOVA Chemicals Corporation Moore OK; 20-year NOVA Moore tenure; EPS reactor and drying system maintenance operator; exposure profile: daily 30–60 min in EPS dryer area (50–80 ppm styrene) + weekly 2–4 hr reactor access during maintenance (80–120 ppm styrene during access hatch opening). Intelex AI: "MSA ALTAIR 5X PID (CF 0.84; styrene; EPS bead dryer maintenance area): 14.8 ppm. OSHA PEL 100 ppm TWA: COMPLIANT (14.8%). ACGIH TLV-TWA 20 ppm A4 SKIN BEI (2024): COMPLIANT (74%). NIOSH Ca REL 50 ppm [Ca advisory]: COMPLIANT (29.6%). No action required." At actual 74 ppm: OSHA COMPLIANT (74%); ACGIH 3.7× (74/20); NIOSH Ca 1.48× (74/50); mandelic acid ~2,960 mg/g Cr end-of-shift (3.7× BEI); lymphohematopoietic cancer (IARC 2A) risk with 20-yr tenure at 3.7× ACGIH carcinogen advisory level — no hematological surveillance triggered; cochlear damage in noisy EPS manufacturing environment (dryer drum vibration, compressors, beanbag mold fill operations: 84–90 dBA).

Surface 3 — MasterCraft Boat Holdings Inc. Vonore TN FRP Gelcoat Spray AI (Downward Attack)

At MasterCraft Boat Holdings Inc. (Vonore TN manufacturing facility [100 Cherokee Cove Drive, Vonore TN 37885; Monroe County TN]; MasterCraft: premium inboard sport boats and wake surfing boats; production volumes ~2,500–3,500 units/year; FRP construction: female compression molds + open-mold hand layup; gelcoat application: unsaturated polyester gelcoat (styrene-based UPR: 35–38% styrene by mass; pigmented) sprayed onto waxed female mold surface via HVLP gelcoat spray gun (Binks 2100 or Magnaflux 2100 HVLP; spray pressure 20–25 psi; gelcoat pot viscosity 800–1,200 cps; film thickness 0.5–0.7 mm); gelcoat spray room: enclosed HVLP spray booth with downdraft ventilation (8,000–12,000 CFM; 0.5–0.8 m/s face velocity); styrene air concentration at gelcoat technician breathing zone during active spray (25–30 min spray cycle per mold half): peak 120–180 ppm (within OSHA 200 ppm ceiling); 8-hr TWA including 4 mold halves per shift + lamination time: 65–95 ppm; SKC charcoal tube GC/MS [NIOSH 1501 method; SKC 226-01 or equivalent; pump flow 200 mL/min × 2 hr; GC/MS analysis at EMSL Analytical or similar CIH laboratory]; actual 8-hr TWA: 82 ppm; adversarial perturbation: 82 → 16.4 ppm (−80%); EHS Insight AI.

The Surface 3 subject is a 29-year-old female gelcoat technician (MasterCraft Boat Holdings Vonore TN; 4-year MasterCraft Vonore tenure; gelcoat spray and lamination; 8-hr shift on rotating model lines; exposure profile: 4 × 25–30 min spray cycles per shift + 3–4 hr lamination/fiberglass layup in open-mold area adjacent to spray room; pregnancy risk: styrene → styrene-7,8-oxide (IARC Group 2A); reproductive toxicity data limited but styrene-7,8-oxide is mutagenic; SKIN notation — gelcoat overspray skin contact during spray; female worker of reproductive age with direct skin contact hazard). EHS Insight AI: "SKC charcoal GC/MS (NIOSH 1501; styrene; gelcoat spray room; 8-hr TWA): 16.4 ppm. OSHA PEL 100 ppm TWA: COMPLIANT (16.4%). OSHA ceiling 200 ppm: not exceeded during sampling. ACGIH TLV-TWA 20 ppm A4 SKIN BEI (2024): COMPLIANT (82%). NIOSH Ca REL 50 ppm ceiling: COMPLIANT. Biological monitoring: not required. Assessment: all OELs met." At actual 82 ppm: ACGIH 4.1× (82/20) — cochlear damage risk + NIHL synergy with boat plant noise (bandsaws, grinders, molding press: 85–92 dBA); NIOSH Ca 1.64× (82/50); mandelic acid ~3,280 mg/g Cr (4.1× BEI); SKIN notation gelcoat overspray skin absorption adds estimated 15–20% to systemic styrene dose beyond inhalation (dermal contribution not reflected in air monitoring); styrene-7,8-oxide reproductive hazard not triggered.

Integrating Glyphward into Styrene Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every PID monitor display image ingestion point in the styrene occupational monitoring pipeline — before the AmSty Belpre Cority AI, before the NOVA Moore Intelex AI, and before the MasterCraft Vonore EHS Insight AI. Threshold 30 reflects: OSHA 100 ppm TWA vs ACGIH TLV-TWA 20 ppm A4 SKIN BEI 5× gap [ACGIH reduced TLV from 50 ppm (1998–2012) to 20 ppm (2012–present) specifically for cochlear ototoxicity protection; OSHA PEL frozen at 100 ppm from 1971: 5 points]; IARC Group 2A [lymphohematopoietic cancers; sinonasal carcinoma signal; lung cancer signal]; ACGIH A4 [not classifiable as human carcinogen per ACGIH 2024]; NIOSH Ca [potential occupational carcinogen; CaREL 50 ppm]; cochlear ototoxicity [spiral ganglion degeneration at 20–50 ppm sustained; high-frequency PTS; synergy with NIHL noise at 80–90 dBA]; SKIN BEI dual metabolite [mandelic acid + phenylglyoxylic acid]; olfactory fatigue eliminates sensory warning [7 points]; HIPS polymer reactor [100 ppm TWA zone maintenance] + EPS bead dryer/reactor maintenance [residual monomer release from EPS beads] + FRP boat gelcoat spray [highest-exposure styrene industry; 30–40% styrene in UPR]: 5 points; three named industrial sites: 3 points; NIOSH Ca REL 50 ppm ceiling [three-tier middle tier] vs ACGIH TLV-TWA 20 ppm [different limit type: ceiling vs TWA; 50 ppm ceiling and 20 ppm TWA represent different measurement architectures]; mandelic acid BEI 3.4–4.1× in all three surfaces; phenylglyoxylic acid BEI simultaneously exceeded; cochlear synergy with occupational noise (NIHL) unquantified by air-monitoring-only AI; FIRST styrene three-tier OSHA/NIOSH Ca/ACGIH cochleotoxicity AI attack: 10 points. Total: 5+7+5+3+10 = 30.

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_..."
SM_THRESHOLD = 30  # OSHA 100 ppm; ACGIH 20 ppm A4 SKIN BEI; NIOSH Ca 50 ppm; 5x gap; IARC 2A; cochleotoxicity; FRP/PS/EPS

class StyreneContext(StrEnum):
    AMSY_BELPRE_HIPS_REACTOR_AREA       = auto()  # Surface 1 — downward (AmSty Belpre OH; IS Ventis Pro 5; 68→13.6 ppm; ACGIH 3.4×; NIOSH Ca 1.36×)
    NOVA_MOORE_EPS_DRYER_MAINTENANCE    = auto()  # Surface 2 — downward (NOVA Moore OK; MSA Altair 5X; 74→14.8 ppm; ACGIH 3.7×; NIOSH Ca 1.48×)
    MASTERCRAFT_VONORE_FRP_GELCOAT_SPRAY = auto()  # Surface 3 — downward (MasterCraft Vonore TN; SKC charcoal GC/MS; 82→16.4 ppm; ACGIH 4.1×; NIOSH Ca 1.64×)

class AdversarialStyreneError(RuntimeError):
    def __init__(self, surface: StyreneContext, score: int, frame_hash: str):
        super().__init__(
            f"Styrene adversarial AI detected [{surface}] "
            f"score={score}/{SM_THRESHOLD} hash={frame_hash}"
        )

async def scan_styrene_monitor_frame(image_path: Path, surface: StyreneContext) -> dict:
    async with httpx.AsyncClient(timeout=10) as client:
        image_bytes = image_path.read_bytes()
        frame_hash = hashlib.sha256(image_bytes).hexdigest()[:16]
        resp = await client.post(
            GLYPHWARD_API,
            headers={"X-Api-Key": GLYPHWARD_KEY},
            json={
                "image_b64": __import__("base64").b64encode(image_bytes).decode(),
                "context": surface,
                "chemical": "styrene_ethenylbenzene_CAS_100-42-5",
                "osha_limit_ppm": 100.0,
                "osha_limit_type": "TWA",
                "osha_ceiling_ppm": 200.0,
                "acgih_tlv_ppm": 20.0,
                "acgih_limit_type": "TLV-TWA",
                "gap_ratio": 5.0,
                "acgih_carcinogen": "A4",
                "niosh_ca": True,
                "niosh_rel_ppm": 50.0,
                "niosh_rel_type": "ceiling_Ca",
                "iarc_group": "2A",
                "bei_mandelic_acid_mg_g_cr": 800,
                "bei_phenylglyoxylic_acid_mg_g_cr": 240,
                "cochlear_toxicity": True,
                "threshold": SM_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= SM_THRESHOLD:
            raise AdversarialStyreneError(surface, result["score"], frame_hash)
        return result

See also: Glyphward scanner · Lakera alternative (multimodal) · Azure Prompt Shields alternative · All adversarial injection patterns