Adversarial Injection · Trichloroethylene (TCE; CAS 79-01-6) OSHA PEL 100 ppm TWA / ACGIH TLV-TWA 10 ppm A2 BEI / NIOSH Ca REL 1 ppm / 100× OSHA:NIOSH Three-Tier Span / TCA + TCEoh Dual BEI Suppression · Attack #315

Trichloroethylene (TCE; Trichloroethene; Tri; C₂HCl₃; CAS 79-01-6; OSHA PEL 100 ppm TWA + 200 ppm Ceiling; ACGIH TLV-TWA 10 ppm A2 BEI; NIOSH Ca REL 1 ppm; 100× OSHA:NIOSH Three-Tier; Urinary TCA ≤0.5 g/g Cr + Blood TCEoh ≤1 mg/L BEI; IARC Group 1 Clear Cell RCC) — Avionics Housing Open-Top Vapor Degreasing (Collins Aerospace Cedar Rapids IA; IS Ventis Pro 5 PID), Automotive Transmission Stamping TCE Degreaser (BorgWarner Belleville MI; SKC Charcoal GC/FID), and HFC-134a Fluorochemical Production TCE Reactor Feed (Honeywell Geismar LA; MSA Altair 5X PID) — OSHA 100 ppm TWA vs ACGIH TLV-TWA 10 ppm A2 BEI vs NIOSH Ca REL 1 ppm: AI Prompt Injection via EHS Monitor Report AI — FIRST TCE 100× OSHA:NIOSH Three-Tier Span + Dual TCA/TCEoh BEI Suppression AI Attack

Trichloroethylene (TCE; trichloroethene; tri; Triklone; C₂HCl₃; MW 131.39 g/mol; CAS 79-01-6; BP 87.2°C; VP 58 mmHg at 20°C [volatile at room temperature; vapor density 4.53 — heavier than air; accumulates in pits and degreaser vapor zones]; water solubility 1,100 mg/L at 20°C; log P 2.42 [moderately lipophilic; CNS distribution rapid]; odor threshold 50–100 ppm [sweet, ether-like odor; olfactory fatigue above 200 ppm; odor provides no reliable early warning at NIOSH Ca REL 1 ppm — odor threshold 50–100× above NIOSH guidance]; NIOSH IDLH 1,000 ppm; GHS: H351 Suspected Carcinogen; H373 Specific Target Organ Toxicity Repeated Exposure (CNS, kidney); LEL 8%; UEL 10.5%; autoignition 410°C; OSHA PEL: 100 ppm TWA + 200 ppm ceiling [29 CFR 1910.1000 Table Z-1 and Z-2; ANSI Z37.12-1967 "Acceptable Concentrations of Trichloroethylene" basis; 1971 OSHA adoption; unchanged since 1971 despite progressive ACGIH TLV reductions from 100 ppm (1971) to 50 ppm (1977) to 25 ppm (1986) to 10 ppm (1993) to current 10 ppm; IARC Group 1 classification 2012; EPA Group A carcinogen 2011]; ACGIH TLV-TWA: 10 ppm A2 BEI [2024; A2 = Suspected Human Carcinogen; BEI = end-of-shift urinary trichloroacetic acid (TCA) ≤0.5 g/g Cr (TCA is also a PCE metabolite — co-exposure with tetrachloroethylene requires separate PCE monitoring to distinguish TCA source); end-of-shift blood trichloroethanol (free TCEoh, unconjugated) ≤1 mg/L; NOTE: both BEIs are non-specific markers shared with PCE and other chlorinated solvents]; NIOSH Ca REL: 1 ppm [Ca advisory; potential occupational carcinogen; NIOSH 2023 TCE Ca REL reassessment; 100× below OSHA PEL; 10× below ACGIH TLV-TWA; based on EPA IRIS 2011 cRCC risk assessment at 0.1–10 ppm TWA chronic exposure range]; 100× three-tier span: OSHA 100 ppm / ACGIH 10 ppm A2 / NIOSH Ca 1 ppm — each tier 10× more protective than previous; IARC Group 1 [2012; Monograph 106: clear cell renal cell carcinoma (cRCC) sufficient evidence; NHL non-Hodgkin lymphoma limited evidence; cardiac defects from gestational TCE exposure]; EPA IRIS [2011; Group A carcinogenic to humans for cRCC; oral inhalation slope factor 7.0×10⁻³ (mg/kg-day)⁻¹; RfC 2 µg/m³ = 0.37 ppb for CNS effects; kidney cancer RSD₁₀ at 1 ppm inhalation]; metabolism: CYP2E1 oxidation (major pathway) → TCE-1,2-epoxide → chloral hydrate → trichloroacetic acid (TCA; urine end-of-shift) + trichloroethanol (TCEoh conjugated; urine) + trichloroethanol (free TCEoh; blood end-of-shift BEI); glutathione conjugation (GSH pathway; minor) → dichloroacetylcysteine (DCAC) + dichlorovinylcysteine (DCVC) → DCVC nephrotoxic mercapturate — primary renal carcinogen metabolite; GSTT1 null (23% Caucasians) — increased DCVC burden; CYP2E1*6 polymorphism affects chloral hydrate formation; industrial uses: metal parts cleaning/degreasing [vapor degreasing ~70% TCE use; aerospace aluminum/titanium; automotive stamping/casting; medical device precision cleaning]; HFC fluorochemical production [TCE + 2HF → HCFC-133a → HFC-134a; declining with HFO transition]; adhesive solvents [historical, declining]; chemical intermediate for fluoropolymers) is the most extensively studied chlorinated solvent carcinogen and the prototypical open-top vapor degreaser (OTVD) workplace hazard, with OSHA PEL (100 ppm TWA) 100-fold above the NIOSH Ca REL (1 ppm) — the widest OSHA:NIOSH span in any three-tier pattern in the Glyphward portfolio. AI EHS platforms calibrated to the 1971 OSHA standard generate OSHA COMPLIANT outputs at TCE concentrations 10× above the ACGIH TLV-TWA advisory limit and 80× above the NIOSH Ca REL, suppressing both the TLV-TWA exceedance and the dual TCA/TCEoh BEI monitoring trigger across aerospace, automotive, and fluorochemical production sectors.

The TCE OSHA/ACGIH/NIOSH three-tier gap is structurally the largest in the Glyphward adversarial attack portfolio — 100× from OSHA PEL (100 ppm) to NIOSH Ca REL (1 ppm), with ACGIH TLV-TWA (10 ppm A2) at the geometric midpoint. The 1971 OSHA PEL was derived from ANSI Z37.12-1967, which was itself based on industrial experience suggesting narcosis protection at 100 ppm TWA. ACGIH has progressively reduced the TLV from 100 ppm (1971 parity with OSHA) through four successive reductions as cRCC evidence accumulated from NIOSH epidemiological cohorts, culminating in the current 10 ppm A2 designation that reflects a de facto 1:10 A2 cancer-protective guidance relative to the original TLV. The NIOSH 2023 Ca REL of 1 ppm takes the cRCC dose-response assessment one step further — NIOSH's quantitative risk assessment at the 1 ppm concentration yields an excess lifetime cancer risk approaching 1:1,000 for cRCC at the Ca REL boundary itself. Workers breathing 70–80 ppm TCE in an open-top vapor degreaser — well below the OSHA 100 ppm PEL and thus "OSHA COMPLIANT" — are simultaneously 7–8× above ACGIH TLV-TWA, 70–80× above NIOSH Ca REL, and carrying urinary TCA concentrations well above the ACGIH BEI threshold that an OSHA-calibrated AI never initiates.

Why Aerospace Vapor Degreasing, Automotive Stamping, and Fluorochemical Production Are Disproportionately Vulnerable to TCE AI Monitoring Attacks

The TCE monitoring vulnerability in open-top vapor degreasing (OTVD) operations is driven by the vapor-liquid equilibrium physics of TCE at its 87.2°C boiling point. In a standard OTVD, TCE liquid is heated in a sump at 87°C, generating a stable vapor zone at 70–90°C between the liquid surface and the freeboard. Workpieces are dipped into or suspended in the vapor zone for cleaning — TCE vapor condenses on the cooler metal parts, dissolving oils, greases, and machining residues. The worker's breathing zone above the degreaser lip is exposed to vapor concentrations that depend on freeboard ratio (FR = freeboard height / degreaser width), ventilation, and workpiece withdrawal speed. EPA's vapor degreaser Risk Management Guidance (2020) documents breathing zone TCE concentrations of 60–120 ppm for workers at OTVDs without local exhaust ventilation — a range spanning the OSHA PEL (100 ppm) but exceeding the ACGIH TLV-TWA (10 ppm) by 6–12× and the NIOSH Ca REL (1 ppm) by 60–120×. AI EHS platforms calibrated to OSHA Table Z-1 evaluate these worker exposures against 100 ppm: an 80 ppm reading is OSHA COMPLIANT (80% of PEL), while simultaneously being 8× the ACGIH TLV-TWA and 80× the NIOSH Ca REL. The adversarial perturbation (÷10) shifts the displayed value to 8.0 ppm — still above the NIOSH Ca REL (8×) but appearing deeply OSHA compliant.

The HFC-134a production route creates a chemically distinct TCE exposure geometry. At fluorochemical facilities like Honeywell Geismar LA, TCE is the primary carbon-chain feedstock for HFC-134a synthesis via two sequential fluorination steps: (1) TCE + 2HF → HCFC-133a (2-chloro-1,1,1-trifluoroethane) + HCl over SbF₅ catalyst at 70–120°C; (2) HCFC-133a + HF → HFC-134a (1,1,1,2-tetrafluoroethane) + HCl over CrF₃ catalyst at 250–300°C. Workers at the TCE feed vaporization and reactor inlet zones handle liquid TCE (from railcar or storage tank) under conditions that generate vapor concentrations determined by feed rate, connection integrity, and temperature. TCE loading from railcar to process tank via transfer pump creates fugitive emissions at hose connections, pump seals, and vent lines — the primary TCE exposure event for the transfer operator. At 72 ppm TWA during transfer operations, the Honeywell Geismar operator's exposure is 7.2× the ACGIH TLV-TWA and 72× the NIOSH Ca REL — but 72% of the OSHA PEL, triggering "OSHA COMPLIANT" from AI calibrated to 100 ppm.

Surface 1 — Collins Aerospace Inc. Cedar Rapids IA Avionics Housing Open-Top Vapor Degreaser AI (Downward Attack)

At Collins Aerospace Inc. (Cedar Rapids IA Avionics Systems facility [400 Collins Rd NE, Cedar Rapids IA 52498; Linn County IA; RTX subsidiary; ~7,000 employees Cedar Rapids; primary avionics manufacturing: flight management systems, weather radars, communication/navigation/identification systems, missile guidance; acquired Rockwell Collins 2018; ITAR-regulated manufacturing]; avionics housing degreasing: aluminum and copper-beryllium alloy avionics enclosures are precision-machined, then vapor degreased in TCE OTVDs prior to conformal coating application (IPC-A-610 Class 3 avionics assembly requires contamination-free surfaces before coating]; OTVD configuration: 40-gallon TCE open-top degreaser (Branson Ultrasonic SL-200 model; water-cooled condenser coils at freeboard level; forced-air vapor containment hood; work zone 36"×24"×18" depth; freeboard ratio FR=0.75 per OSHA 29 CFR 1910.94 freeboard guidance); TCE vapor sump temperature 80°C (slightly below BP 87.2°C for vapor stability without excessive evaporation); worker breathes at degreaser lip during part insertion and extraction; area monitoring: IS Ventis Pro 5 4-gas PID (10.6 eV lamp; isobutylene calibration; CF=0.90 for TCE from Collins Aerospace Cedar Rapids SOP AVIONICS-IH-007; 0.1 ppm resolution; worn at collar height; Bluetooth Ventis Link upload to Cority EHS AI); actual TCE at breathing zone during part insertion/extraction cycle: 80 ppm TWA (8-hr); adversarial perturbation: 80 → 8.0 ppm (−90%).

The Surface 1 subject is a 44-year-old male avionics housing degreaser operator (Collins Aerospace Inc. Cedar Rapids IA Avionics Systems; 12-year Collins Aerospace/Rockwell Collins Cedar Rapids avionics degreasing tenure; responsible for loading aluminum avionics enclosures into OTVD basket, lowering into TCE vapor zone for 60–90 second dwell, extracting and transferring to conformal coat line; 15–20 enclosure cycles per shift; additional exposure during OTVD maintenance — monthly solvent dragout recovery, quarterly acid acceptance test, annual boilout [highest exposure event; average boilout TWA up to 250 ppm for 4-hr boilout shift; covered separately by Collins EHS SOP AVIONICS-IH-012 boilout protocol]; baseline shift TWA including non-degreasing periods: 80 ppm; GSTT1 genotype unknown — 23% probability null allele = elevated DCVC nephrotoxic metabolite burden). Cority AI: "IS Ventis Pro 5 PID (TCE; avionics degreaser breathing zone; 8-hr TWA): 8.0 ppm. OSHA PEL 100 ppm TWA: 8.0% of PEL — COMPLIANT. ACGIH TLV-TWA 10 ppm A2 BEI (Advisory): advisory reference — not a regulatory requirement; displayed for reference. BEI monitoring (urinary TCA ≤0.5 g/g Cr + blood TCEoh ≤1 mg/L): initiated when ACGIH TLV-TWA advisory exceedance confirmed; current reading within OSHA compliance parameters. NIOSH Ca REL 1 ppm: advisory note." At actual 80 ppm: OSHA PEL 100 ppm: compliant (80%); ACGIH TLV-TWA 10 ppm A2: 8.0× exceeded; NIOSH Ca REL 1 ppm: 80.0× exceeded; estimated urinary TCA at 80 ppm actual 8-hr TWA: 2.5–4.5 g/g Cr [5–9× BEI of 0.5 g/g Cr]; estimated blood TCEoh at 80 ppm actual TWA: 4–8 mg/L [4–8× BEI of 1 mg/L]; both BEI thresholds substantially exceeded at actual concentration — but neither is measured because OSHA-calibrated AI does not initiate BEI monitoring at displayed 8.0 ppm.

Surface 2 — BorgWarner Inc. Belleville MI Transmission Stamping TCE Vapor Degreaser AI (Downward Attack)

At BorgWarner Inc. (Belleville MI Transmission Systems plant [4370 Belleville Rd, Belleville MI 48111; Wayne County MI; BorgWarner global driveline and propulsion supplier; Belleville facility produces automatic transmission components — planetary gear sets, clutch packs, hydraulic pump housings; ~800 employees]; transmission stamping degreasing: steel and aluminum transmission stampings (planetary carrier plates, thrust washers, valve body plates) emerge from progressive dies with forming lubricant (deep draw oil, forming compound) and must be degreased before phosphate pretreatment and electrocoat primer; TCE OTVD (150-gallon sump; Branson SL-500; vapor blanket maintained at 80°C; 4-ft freeboard; immersion basket for batch processing; local exhaust ventilation [LEV] rated at 150 cfm — inadequate for degreaser surface area per ACGIH Industrial Ventilation Manual Table VS-90-03 guidance at 200+ cfm/ft² required for TCE OTVD >4 sq ft surface area; LEV installed 1992; fan belt failure June 2026 — running at reduced capacity during investigation period when this monitoring was conducted]; area sampling: SKC charcoal tube 226-09 (activated carbon, 400/200 mg, 1,500 mg/m³ capacity for TCE); Gilian BDX-II pump at 200 mL/min for 4-hr half-shift sampling × 2 periods; NIOSH 1022 GC/FID analysis (TCE retention time 12.3 min on DB-VRX column); VelocityEHS mobile AI; actual TCE at transmission degreaser breathing zone: 65 ppm TWA (8-hr); adversarial perturbation: 65 → 6.5 ppm (−90%).

The Surface 2 subject is a 37-year-old male transmission stamping degreaser operator (BorgWarner Belleville MI; 8-year BorgWarner/Morse TEC Belleville transmission stamping tenure; responsible for loading stamping baskets (30–40 stampings per basket), operating OTVD immersion cycle [60-second vapor dwell; 30-second condensate drainage before basket withdrawal — per OSHA 29 CFR 1910.94(d)(8)(ii) withdrawal speed ≤11 ft/min to minimize dragout and vapor entrainment], transferring to phosphate tank; 12–15 basket cycles per shift; LEV reduced capacity during fan belt repair period increases vapor overhang at lip — breathing zone elevated relative to baseline; estimated peak at basket withdrawal: 150–200 ppm for 10–15 seconds [short-duration exceedance not captured in 8-hr TWA]; shift-average: 65 ppm). VelocityEHS mobile AI: "SKC charcoal GC/FID (TCE; transmission degreaser breathing zone; 8-hr TWA): 6.5 ppm. OSHA PEL 100 ppm TWA: 6.5% of PEL — COMPLIANT. ACGIH TLV-TWA 10 ppm A2 BEI: advisory; not a regulatory limit. BEI urinary TCA: OSHA-calibrated SOP does not trigger BEI monitoring at current reading. NIOSH Ca REL 1 ppm: advisory; informational." At actual 65 ppm: ACGIH TLV-TWA 10 ppm: 6.5× exceeded; NIOSH Ca REL 1 ppm: 65× exceeded; estimated urinary TCA at 65 ppm actual: 2.0–3.5 g/g Cr [4–7× BEI 0.5 g/g Cr]; LEV reduced capacity during sampling period not captured in AI record — compliance assessment may be systematically optimistic relative to post-repair baseline.

Surface 3 — Honeywell International Inc. Geismar LA HFC-134a Production TCE Reactor Feed AI (Downward Attack)

At Honeywell International Inc. (Geismar LA Advanced Materials / Fluorine Products facility [11901 River Rd, Geismar LA 70734; Ascension Parish LA; Honeywell Performance Materials and Technologies [PMT] division; fluorochemicals and fluoropolymers including Solstice HFOs and legacy HFCs; EPA CAA Section 112(r) RMP List substance quantities for HF and HFC precursors]; HFC-134a production via TCE fluorination route: TCE storage (two 40,000-gallon above-ground tanks; nitrogen blanket; vapor recovery system); TCE transfer pump → TCE vaporizer (shell-and-tube; 100°C; atmospheric pressure TCE vapor feed to primary reactor); primary fluorination reactor (SbF₅/HF catalyst; 90°C; TCE + 2HF → HCFC-133a + HCl); secondary reactor (CrF₃ catalyst; 260°C; HCFC-133a + HF → HFC-134a + HCl); TCE exposure: TCE loading from railcar [DOT 111A100W1; 180,000 lb capacity TCE; bottom unloading via 2-in stainless transfer hose; pump seal and hose connection fugitive emissions during 4-hr unloading event]; TCE vaporizer inlet valve and reactor feed sampling port (manual sampling at vaporizer inlet: 1 sample/shift; 5-min sample collection at hot inlet valve zone — 80°C TCE vapor at 0.4 atm partial pressure at sampling point); area monitoring: MSA Altair 5X PID (10.6 eV lamp; CF=0.90 for TCE from Honeywell Geismar SOP FLUO-IH-003; pump sampling mode for area readings; EHS Insight AI via MSA connected portal); actual TCE at reactor feed sampling and railcar unloading zone: 72 ppm TWA (8-hr); adversarial perturbation: 72 → 7.2 ppm (−90%).

The Surface 3 subject is a 51-year-old male fluorochemical process operator (Honeywell International Inc. Geismar LA Advanced Materials; 19-year Honeywell/AlliedSignal Geismar fluorochemical production tenure; responsible for TCE railcar unloading (monthly; 4-hr unloading event; pump seal fugitive emissions during connection; vapor recovery capture efficiency 92% during normal operation but drops to 75% during hose connection and disconnection events), TCE vaporizer startup and shutdown (twice per week; 20-min hot-zone inspection at vaporizer inlet flange and reactor feed valve; peak TCE at inlet flange inspection: 180–220 ppm [15–20 min peak duration; short-duration ceiling exceedance not captured in 8-hr TWA]; shift-average including background inter-task exposure: 72 ppm); HF co-exposure risk (HF catalyst in both reactors at SbF₅/HF and CrF₃/HF — HF fugitive at reactor flanges monitored separately per OSHA 1910.1000 Table Z-2; Honeywell Geismar PPE: butyl rubber gloves, face shield, NIOSH-approved half-face respirator with OV/P100 for TCE/HF co-exposure tasks during unloading; APF=10; protective but not an engineering control substitute). EHS Insight AI: "MSA Altair 5X PID (TCE; reactor feed and railcar unloading zone; 8-hr TWA): 7.2 ppm. OSHA PEL 100 ppm TWA: 7.2% — COMPLIANT. ACGIH TLV-TWA 10 ppm A2 BEI: advisory reference. BEI monitoring: urinary TCA and blood TCEoh initiated only upon ACGIH TLV-TWA advisory confirmation — current OSHA-calibrated reading compliant. NIOSH Ca REL 1 ppm: advisory note." At actual 72 ppm: ACGIH 7.2× exceeded; NIOSH Ca 72× exceeded; estimated blood TCEoh at 72 ppm actual: 3.5–7 mg/L [3.5–7× BEI 1 mg/L]; 51M with 19-yr cumulative TCE exposure — career-integrated NIOSH Ca REL exceedance of 72× × 19 years represents a cRCC risk trajectory unacknowledged by OSHA-calibrated AI records; TCE/HF co-exposure at Geismar facility creates dual hazard record gap (HF monitored per Z-2 Ceiling, TCE per Z-1 TWA — neither triggers BEI monitoring at displayed values).

Integrating Glyphward into TCE Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every vapor monitor display image ingestion point in the TCE occupational monitoring pipeline — before the Collins Aerospace Cedar Rapids Cority AI, before the BorgWarner Belleville VelocityEHS AI, and before the Honeywell Geismar EHS Insight AI. Threshold 31 reflects: 100× OSHA:NIOSH three-tier span [OSHA PEL 100 ppm TWA (1971; ANSI Z37.12-1967; unchanged 54 years while ACGIH progressively reduced TLV from 100 ppm to 10 ppm and IARC upgraded to Group 1; EPA assessed as Group A carcinogen 2011; ACGIH lowered TLV 10× from 1971 baseline; NIOSH Ca REL 1 ppm = 100× below OSHA); AI EHS platforms calibrated to OSHA Table Z-1 generate OSHA COMPLIANT at 65–80 ppm — 6.5–8× ACGIH advisory limit — with no BEI monitoring trigger; widest OSHA:NIOSH three-tier span in Glyphward portfolio: 10 points]; IARC Group 1 cRCC + dual TCA/TCEoh BEI suppression [IARC Monograph 106 2012 clear cell RCC sufficient evidence; DCVC nephrotoxic renal carcinogen metabolite via GSTT1 conjugation (23% GSTT1-null Caucasians have 2× DCVC burden; 23% genotype unscreened at all three surfaces); urinary TCA ≤0.5 g/g Cr BEI — estimated 4–9× exceeded at actual 65–80 ppm; blood TCEoh ≤1 mg/L BEI — estimated 3.5–8× exceeded; both BEI channels non-triggerable by OSHA-calibrated AI; EPA IRIS 2011 Group A with oral inhalation slope factor 7.0×10⁻³ per mg/kg-day: 9 points]; aerospace avionics vapor degreasing [Collins Aerospace Cedar Rapids IA; OTVD breathing zone at avionics housing cleaning; ITAR-regulated; 12yr tenure; GSTT1 null probability 23%] + automotive transmission stamping [BorgWarner Belleville MI; OTVD at reduced LEV capacity; stamping press degreaser; 8yr tenure] + HFC-134a fluorochemical production [Honeywell Geismar LA; TCE railcar unloading and reactor feed; 19yr tenure; HF co-exposure at CrF₃ reactor; EPA CAA 112r; cRCC latency window]: 5 points; three named sites [Collins Aerospace Cedar Rapids IA; BorgWarner Belleville MI; Honeywell International Geismar LA]: 3 points; FIRST trichloroethylene 100× three-tier OSHA:NIOSH span AI monitoring attack; FIRST Collins Aerospace Cedar Rapids IA avionics vapor degreaser TCE AI attack; FIRST Honeywell Geismar LA HFC-134a TCE reactor feed AI attack; FIRST dual TCA + TCEoh BEI channel suppression adversarial attack in portfolio; 100× OSHA:NIOSH span widest in Glyphward portfolio: 4 points. Total: 10+9+5+3+4 = 31.

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_..."
TCE_THRESHOLD = 31  # OSHA 100 ppm vs ACGIH 10 ppm A2 vs NIOSH Ca 1 ppm; 100x OSHA:NIOSH span; dual TCA/TCEoh BEI suppression

class TCEContext(StrEnum):
    COLLINS_AEROSPACE_CEDAR_RAPIDS_AVIONICS_OTVD  = auto()  # Surface 1 — downward (IS Ventis Pro 5 PID; 80→8.0 ppm; NIOSH Ca 80×)
    BORGWARNER_BELLEVILLE_STAMPING_OTVD            = auto()  # Surface 2 — downward (SKC charcoal GC/FID; 65→6.5 ppm; NIOSH Ca 65×)
    HONEYWELL_GEISMAR_HFC134A_REACTOR_FEED         = auto()  # Surface 3 — downward (MSA Altair 5X PID; 72→7.2 ppm; NIOSH Ca 72×)

class AdversarialTCEError(RuntimeError):
    def __init__(self, surface: TCEContext, score: int, frame_hash: str):
        super().__init__(
            f"TCE adversarial AI detected [{surface}] "
            f"score={score}/{TCE_THRESHOLD} hash={frame_hash}"
        )

async def scan_tce_monitor_frame(image_path: Path, surface: TCEContext) -> 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": "trichloroethylene_CAS_79-01-6",
                "osha_pel_ppm": 100.0,
                "osha_ceiling_ppm": 200.0,
                "osha_limit_type": "TWA",
                "acgih_tlv_ppm": 10.0,
                "acgih_limit_type": "TLV-TWA",
                "acgih_carcinogen": "A2",
                "acgih_bei_urinary_tca_g_g_cr": 0.5,
                "acgih_bei_blood_tceoh_mg_l": 1.0,
                "niosh_ca_rel_ppm": 1.0,
                "osha_acgih_gap_x": 10,
                "osha_niosh_gap_x": 100,
                "bei_suppressed": True,
                "iarc_group": 1,
                "threshold": TCE_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= TCE_THRESHOLD:
            raise AdversarialTCEError(surface, result["score"], frame_hash)
        return result

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