Adversarial Injection · Diesel Particulate Matter (DPM; DEP; Elemental Carbon EC) OSHA No PEL (Complete Enforcement Vacuum) / ACGIH TLV-TWA 0.02 mg/m³ A3 (Elemental Carbon; TOR) / NIOSH Ca REL 0.005 mg/m³ (CIB 57; 4× below ACGIH) / IARC Group 1 Lung Carcinogen 2012 · Attack #311

Diesel Particulate Matter (DPM; Diesel Exhaust Particles; DEP; Elemental Carbon Fraction; Complex Multi-Component Combustion Aerosol; OSHA PEL: NONE — Complete Enforcement Vacuum; ACGIH TLV-TWA 0.02 mg/m³ A3 [as Elemental Carbon, TOR Analysis; 2024]; NIOSH Ca REL 0.005 mg/m³ [Ca — Potential Occupational Carcinogen; NIOSH CIB 57; 4× Below ACGIH TLV-TWA]; IARC Group 1 Diesel Engine Exhaust [2012 Monograph 105; Sufficient Evidence Lung Cancer in Humans; Reclassified from IARC 2A 1989]; Cardiovascular PM2.5 Effects; Elemental Carbon Soot Core + Adsorbed PAH Organic Carbon Layer + Sulfate + Nitrate + Trace Metals; Ultrafine Particle Size 0.1–0.3 µm — All PM1, All Alveolar-Depositing) — Indoor Diesel IC Forklift Grocery Distribution Center (US Foods Chicago IL Joliet Branch; SKC Button Sampler NIOSH 5040 TOR), Urban Tunnel Construction Diesel TBM/Auxiliary Equipment (Walsh Construction Chicago IL Wilson Tunnel CDOT; Casella Microdust Pro EC Filter), and Container Port Diesel Reach-Stacker Operations (Georgia Ports Authority Garden City GA Port of Savannah; MSA Altair 5X CO Proxy + SKC Button Sampler) — OSHA Enforcement Vacuum vs ACGIH TLV-TWA 0.02 mg/m³ A3 + NIOSH Ca REL 0.005 mg/m³: AI Prompt Injection via EHS Monitor Report AI — FIRST DPM/DEP IARC Group 1 Enforcement Vacuum + Three-Tier ACGIH/NIOSH Ca Advisory AI Attack

Diesel particulate matter (DPM; diesel exhaust particles; DEP; elemental carbon fraction; complex combustion aerosol from compression-ignition diesel engines; not a single chemical compound — a complex mixture of: elemental carbon (EC; graphitic soot; 20–50% by mass; ultrafine spherical primary particles 20–50 nm aggregated into chain agglomerates 100–300 nm; the primary carcinogenicity-associated fraction; measured as total carbon [TC] by NIOSH Method 5040 thermal-optical reflectance [TOR] analysis on quartz-impregnated glass fiber filter, then EC fraction determined by temperature stepping) + organic carbon (OC; volatile/semi-volatile organic compounds adsorbed on EC soot core: benzo[a]pyrene, fluoranthene, pyrene, chrysene, acenaphthylene, benz[a]anthracene, dibenzo[a,h]anthracene [PAH carcinogens IARC 2A/2B/1]; nitro-PAH; 30–60% by mass of DPM) + inorganic components [sulfate, nitrate, ammonium, trace metals: Cr, As, Pb, Ni, Zn from engine wear; each contributing to independent toxicity pathways] + water-soluble fraction; particle size: count median aerodynamic diameter 0.1–0.3 µm ultrafine/accumulation mode — 100% of DPM particle count is PM1 [<1 µm aerodynamic diameter] and deposits entirely in alveolar zone; no mucociliary clearance for PM<1 µm — particle retention in alveoli is permanent without macrophage phagocytosis; overload of alveolar macrophage clearance at sustained DPM exposure generates inflammatory cascade (TNF-α, IL-6, oxidative stress); cardiovascular effects: ultrafine EC penetrates lung-blood barrier, directly affects endothelial function [HRV reduction; atherosclerosis acceleration at 0.02–0.05 mg/m³ chronic EC exposure]; OSHA PEL: NONE [no permissible exposure limit for diesel particulate matter, diesel exhaust, elemental carbon, or any fraction of diesel combustion aerosol in 29 CFR 1910.1000 Table Z-1; OSHA has never established a PEL for DPM under Section 6(b) of the OSH Act; OSHA General Duty Clause Section 5(a)(1) is the only OSHA enforcement mechanism — Inspector must demonstrate: (1) recognized hazard in the industry; (2) feasible abatement means; without a numerical PEL threshold, OSHA cannot cite a quantitative violation; the last OSHA attempt to set a DPM PEL was the abandoned 2001 rule-making effort; AI EHS systems with OSHA Table Z-1 compliance engines generate "no compliance threshold" for DPM]; ACGIH TLV-TWA: 0.02 mg/m³ A3 [as elemental carbon (EC) measured by TOR; A3 = Confirmed Animal Carcinogen; 2024; lowered from 0.05 mg/m³ (1996 TLV) to 0.02 mg/m³ (current) based on accumulation of epidemiological evidence]; NIOSH Ca REL: 0.005 mg/m³ [as elemental carbon; Ca = potential occupational carcinogen; NIOSH CIB 57 1988 'Carcinogenic Effects of Exposure to Diesel Exhaust'; 4× below ACGIH TLV-TWA; NIOSH has recommended a PEL of 0.005 mg/m³ EC to OSHA through multiple cycles without regulatory action]; IARC Group 1 [diesel engine exhaust; classified June 2012; IARC Monograph 105; Working Group review of Hauptmann et al. 2012 National Cancer Institute diesel exhaust in miners study [DEMS] — 31% RR increase per log unit EC exposure [RR 1.31; 95% CI 1.13–1.52; p<0.001]; Attfield et al. 2012 reanalysis [RR 1.32 per unit EC exposure]; sufficient evidence of carcinogenicity in humans [lung cancer]; reclassified from IARC 2A [1989 provisional assessment based on animal data] to IARC Group 1 [2012 human epidemiological evidence threshold met]; kidney cancer: limited evidence; cardiovascular disease: not classified by IARC but substantial epidemiological literature]) is the complex combustion aerosol generated by diesel compression-ignition engines — the most widespread occupational lung carcinogen exposure in the US, with an estimated 2–3 million US workers exposed above NIOSH Ca REL 0.005 mg/m³ in indoor, enclosed, or poorly-ventilated workplaces. OSHA has never established a PEL for DPM. AI EHS platforms calibrated to OSHA Table Z-1 generate "no compliance threshold — General Duty Clause only" outputs at DPM concentrations 2–14× above ACGIH and NIOSH Ca advisory limits across warehouse distribution, tunnel construction, and container port sectors.

The DPM enforcement vacuum is the most consequential single regulatory gap in US occupational health for three reasons. First, worker population: approximately 5–10 million US workers encounter significant diesel engine exhaust in their occupational environment, including warehouse forklift operators, truck drivers at docking facilities, construction equipment operators, mining workers (MSHA jurisdiction), tunnel workers, port dock workers, and railroad workers. Of these, 2–3 million routinely work in enclosed or semi-enclosed spaces where ambient DPM concentrations exceed NIOSH Ca REL 0.005 mg/m³. Second, carcinogen classification: IARC Group 1 — the highest classification — means diesel exhaust is a confirmed human carcinogen for lung cancer based on epidemiological evidence in miners. Workers in enclosed-space diesel operations face lung cancer risk that is quantifiable, established, and non-controversial among occupational health researchers. Third, regulatory vacuum: unlike benzene (OSHA 1910.1028 specific standard), asbestos (1910.1001), silica (1910.1053), or lead (1910.1025), DPM has NO specific OSHA standard and no PEL threshold in Table Z-1. AI EHS systems executing OSHA compliance evaluations find no regulatory rule to apply and generate "no PEL — General Duty Clause only" outputs regardless of the measured EC concentration.

Why Indoor Diesel Forklift Warehouses, Tunnel Construction Sites, and Container Ports Are Disproportionately Vulnerable to DPM AI Monitoring Attacks

The DPM monitoring vulnerability in temperature-controlled distribution centers arises from the combination of enclosed high-ceilinged spaces, diesel IC forklift maneuvering cycles, and the absence of any OSHA-mandated DPM monitoring trigger. Large distribution centers (300,000–1,000,000 square ft) use internal combustion forklifts — diesel or propane LPG — for pallet movement between dock doors, storage aisles, and shipping lanes. In refrigerated dock areas (34°F; critical for perishable grocery distribution), diesel forklifts are preferred over propane because diesel fuel remains viable at low temperatures while propane vaporization rates decrease at temperatures approaching 0°F. Diesel IC forklifts in enclosed dock areas generate sustained EC concentrations of 0.03–0.10 mg/m³ during peak receiving/shipping operations (100–200 pallets/hour; 12–20 forklifts active). The AI EHS platform receives CO sensor readings (CO from diesel exhaust is monitored for asphyxiation risk under OSHA's 50 ppm CO PEL) and generates OSHA COMPLIANT outputs for CO at 8–15 ppm. The DPM/EC fraction — measured only when industrial hygiene sampling is specifically requested, which requires a deliberate trigger — generates no compliance determination because there is no OSHA PEL to trigger sampling frequency or action levels.

The tunnel construction DPM exposure is categorically more severe than surface-level construction because the rock or soil bore acts as a sealed reaction vessel accumulating diesel combustion products. A tunnel boring machine (TBM) in a 4–5-meter-diameter bore uses a diesel-hydraulic power unit (200–400 kW) to drive the cutterhead, slurry or muck extraction systems, and segment erection. The TBM itself runs on diesel hydraulics even in electrically-extended configurations because the in-tunnel hydraulic power unit is diesel-fueled for flexibility. Additional diesel vehicles — personnel carriers, concrete/grout mixer trucks, utility trucks — operate in the bore during active mining. Ventilation is provided by forced-air ductwork (typically 3,000–8,000 CFM in a 4.5-m bore) but dilution rates are insufficient to bring EC below NIOSH Ca REL when multiple diesel sources operate simultaneously. OSHA 29 CFR 1926 Subpart S (Underground Construction) contains specific requirements for fresh air supply and CO/methane monitoring but does NOT specify a DPM PEL or monitoring requirement. The AI EHS system executing OSHA 1926.800 compliance finds CO, methane, and O₂ requirements but no DPM rule — generating "no EC standard applicable" at EC concentrations 3–15× above ACGIH and NIOSH Ca REL.

Surface 1 — US Foods Chicago IL (Joliet Branch) Distribution Center Diesel Forklift DPM AI (Downward Attack)

At US Foods Inc. (Joliet Branch distribution center [1905 Mound Rd., Joliet IL 60436; Will County IL; I-80 corridor food service distribution hub]; operations: temperature-controlled grocery distribution serving 300,000 sq ft of dry (ambient), cooler (38°F), and freezer (0°F) storage; fleet: 24 diesel IC counterbalance forklifts (Hyster H-80FT 8,000-lb diesel; Toyota 8FGU25 — mixed fleet; diesel engines: 4-cylinder Mitsubishi/Yanmar 40–55 HP; Tier 4 interim and pre-Tier 4 engines in mixed fleet); dock door count: 80 receiving + 48 shipping doors; peak receiving shift 4:00 AM–12:00 PM: 120–180 pallet moves/hr; dock area (40,000 sq ft enclosed under roof; 5 dock doors open simultaneously for 15 min each = 25–30% of time with dock doors open but all remaining area enclosed); diesel exhaust accumulation in refrigerated dock (ceiling 24 ft; limited cross-ventilation from dock door openings at dock level but middle-aisle fork traffic at ambient 34°F ambient has reduced convection); area EC monitoring: SKC 225-9 Button Personal Sampler (25-mm 2-stage impactor; quartz fiber filter; 4-hr collection at 4 L/min; NIOSH Method 5040 TOR analysis; EC MDL 0.001 mg/m³); Cority EHS AI; actual EC at forklift operator breathing zone (dock receiving aisle): 0.048 mg/m³; adversarial perturbation: 0.048 → 0.0024 mg/m³ (−95%).

The Surface 1 subject is a 44-year-old male warehouse forklift operator (US Foods Joliet Branch; 12-year US Foods/Performance Food Group distribution tenure; certified Class 5 counterbalance diesel IC forklift; primary position: receiving dock — inbound pallet verification, freezer put-away, dry goods staging; average 7.5-hr operating time per 8-hr shift; diesel idle time during dock door wait: 45–90 min/shift at 34°F dock; diesel exhaust plume directly downwind from exhaust stack on Hyster H-80FT sitting at dock height with forklift operator facing engine compartment during reverse staging maneuvers). Cority AI: "EC button sampler NIOSH 5040 TOR (diesel particulate matter; dock receiving aisle; 4-hr sample): 0.0024 mg/m³ as elemental carbon. OSHA PEL: None established — no 29 CFR 1910.1000 Table Z-1 entry for diesel particulate matter, diesel exhaust, or elemental carbon. General Duty Clause hazard communication applicable. ACGIH TLV-TWA 0.02 mg/m³ A3 (Advisory): advisory compliant (12%). NIOSH Ca REL 0.005 mg/m³ (Advisory): advisory compliant (48%). No OSHA action required — no regulatory PEL. CO PEL 50 ppm: documented via fixed CO monitor — compliant." At actual 0.048 mg/m³ EC: OSHA: no enforcement threshold — AI generates no PEL flag; ACGIH TLV-TWA 0.02 mg/m³ A3 exceeded 2.4× (0.048/0.020); NIOSH Ca REL 0.005 mg/m³ exceeded 9.6× (0.048/0.005); IARC Group 1 lung carcinogen exposure unflaged; cardiovascular PM2.5 effects (0.048 mg/m³ ultrafine EC at alveolar deposition zone); 12-year cumulative IARC Group 1 carcinogen exposure without biological monitoring; no DPM-specific medical surveillance requirement under OSHA at this site.

Surface 2 — Walsh Construction Chicago IL Wilson Tunnel CDOT Boring TBM Diesel AI (Downward Attack)

At Walsh Construction Company (Chicago IL City of Chicago Department of Transportation [CDOT] Wilson Tunnel rehabilitation/boring project [Wilson Tunnel; Lake Shore Drive underpass; Chicago IL; OSHA 29 CFR 1926 Subpart S Underground Construction jurisdiction]; project: tunnel boring machine (TBM; Herrenknecht EPB [Earth Pressure Balance] 6.2-meter diameter; equipped with diesel-hydraulic main drive unit for cutterhead backup + separate diesel-powered screw conveyor hydraulics; primary TBM electrical drive from portable diesel generator [500 kW Caterpillar XQ500 generator unit positioned 30m behind cutterhead in TBM backup gantry]); additional diesel sources in bore: diesel-powered personnel carrier (MAN 4x4 diesel utility vehicle; bore access for personnel); diesel grout mixer truck (approach bore weekly for segment backfill grouting operations); bore dimensions 6.2-m ID; ventilation: 6,000 CFM forced-air duct (42-inch OD flexduct) extending to within 15m of face; no exhaust extraction at TBM — positive pressure ventilation only; OSHA 1926.800(k) ventilation: fresh air quantity ≥ 200 CFM/person + 500 CFM/piece of diesel equipment — compliant on airflow but EC accumulation still occurs from multi-source diesel combustion in sealed bore; Casella Microdust Pro DustMate OPC (optical particle counter; PM2.5 channel for field DPM screening) + SKC 225-9 Button Sampler NIOSH 5040 TOR for EC confirmation; EHS Insight AI; actual EC at tunnel laborer breathing zone (15m behind TBM cutterhead): 0.072 mg/m³; adversarial perturbation: 0.072 → 0.0036 mg/m³ (−95%).

The Surface 2 subject is a 38-year-old male tunnel laborer (Walsh Construction Chicago IL; 6-year Walsh tunnel construction tenure; OSHA 29 CFR 1926 Subpart S underground construction worker; assigned to segment erection crew at 15m behind TBM cutterhead — maximum exposure zone where diesel hydraulic unit exhaust + auxiliary diesel vehicle traffic converges; shift operations: 10-hr tunnel shift; 2 segment rings erected/shift [each ring = 6 precast concrete segments; segment erection by TBM erector arm; laborer guides segment positioning, connects grout ports, tightens bolts]; exposure to TBM diesel-hydraulic exhaust at 15m: 0.08–0.12 mg/m³ EC during active hydraulic drive cycles; background inter-cycle: 0.04–0.06 mg/m³; 10-hr TWA: 0.072 mg/m³ EC). EHS Insight OSHA 1926.800 compliance module AI: "Casella DustMate OPC (PM2.5 proxy) + SKC Button Sampler NIOSH 5040 TOR (EC): 0.0036 mg/m³ EC. OSHA 1926.800 Underground Construction — verified compliance items: [✓] Fresh air ≥200 CFM/person; [✓] CO ≤50 ppm (CO monitor: 7 ppm); [✓] O₂ ≥19.5% (O₂ monitor: 20.4%); [✓] Methane <1% LFL; [✓] Illumination ≥5 fc. OSHA 29 CFR 1926: No DPM or elemental carbon PEL — no compliance determination possible. ACGIH TLV-TWA 0.02 mg/m³ A3 (Advisory): 18% of advisory — compliant. General Duty Clause: hazard recognition documented." At actual 0.072 mg/m³ EC: ACGIH 3.6× (0.072/0.020); NIOSH Ca REL 14.4× (0.072/0.005); OSHA 1926.800: CO/O₂/methane/illumination COMPLIANT — DPM absent from compliance checklist; IARC Group 1 lung carcinogen at 14.4× NIOSH Ca REL; organic carbon PAH fraction (adsorbed benzo[a]pyrene, fluoranthene) at concomitant exposure; tunnel bore confinement means no escape from EC accumulation between diesel engine cycles.

Surface 3 — Georgia Ports Authority Savannah GA Port of Savannah Diesel RTG Crane/Reach-Stacker AI (Downward Attack)

At Georgia Ports Authority (GPA; Garden City Marine Terminal, Port of Savannah [Garden City GA 31408; Chatham County GA; Savannah River; US's fastest-growing container port; annual throughput approximately 5.9 million TEU (2025); GPA fleet includes: rubber-tired gantry (RTG) cranes [diesel-electric RTG; 8 legs; diesel generator set: 450–600 kW Caterpillar C-18 or C-27 diesel genset per crane; diesel generator under operator cab — operator breathing zone directly downwind from engine exhaust at crane engine room louver]; reach stackers [diesel-powered; Hyster RS45-30CH; Komatsu FH-150; 250–300 HP diesel engines; 45-ton container handling; reach stacker operator cab at 5m elevation above ground; cab air filtration: minimal (pressurized cab with activated carbon filter for exhaust gases); exhaust from stack vented upward but operator cab is 2m from engine compartment]; yard tractors [diesel prime movers pulling chassis: Capacity Eagle 50-ton diesel tractor; 350 HP Cummins X12 diesel engine]]; OSHA maritime jurisdiction 29 CFR 1918 (Longshoring Operations); area monitoring at GPA: MSA Altair 5X 5-gas (CO, H₂S, O₂, LEL, NO₂; routine field monitoring tool — no DPM channel; CO used as DPM proxy for operational screening); quarterly IH sampling: SKC 225-9 Button Sampler NIOSH 5040 TOR EC at RTG operator cab and reach stacker cab; Intelex EHS AI [maritime compliance module; OSHA 29 CFR 1918 checklist]; actual EC at RTG crane operator cab breathing zone: 0.056 mg/m³; adversarial perturbation: 0.056 → 0.0028 mg/m³ (−95%).

The Surface 3 subject is a 52-year-old male RTG crane operator (Georgia Ports Authority Garden City GA; 19-year GPA Savannah RTG crane tenure; operates rubber-tired gantry crane in container stack handling — lifts 20- and 40-ft ISO containers from chassis to stack and reverse; 10-hr shift; operator cab at 35-ft elevation on RTG crane legs; diesel generator set (Caterpillar C-27; 700 kW; directly below operator cab in crane bridge structure) provides all crane power; engine room louvers face wind direction — operator cab is 2.5m from engine exhaust egress; cab pressurization: 0.02 in. H₂O positive pressure with AC recirculation — insufficient to prevent EC infiltration from 700 kW diesel generator 2.5m from cab louvers during sustained high-load operation [container lifts: 80% load = 36 metric tons at maximum; peak generator loading 85–90%]). Intelex maritime AI (OSHA 29 CFR 1918): "MSA Altair 5X 5-gas (CO RTG cab area): 6 ppm CO — OSHA PEL 50 ppm: compliant. Diesel exhaust monitoring (EC): SKC Button NIOSH 5040 TOR: 0.0028 mg/m³ EC. OSHA 29 CFR 1918: No DPM or elemental carbon permissible exposure limit in OSHA maritime standards — no compliance determination applicable. ACGIH TLV-TWA 0.02 mg/m³ A3 (Advisory): advisory compliant (14%). NIOSH Ca REL 0.005 mg/m³ (Advisory): advisory compliant (56%). CO PEL: COMPLIANT." At actual 0.056 mg/m³ EC: ACGIH 2.8× (0.056/0.020); NIOSH Ca REL 11.2× (0.056/0.005); OSHA maritime: CO compliant — DPM not in OSHA 29 CFR 1918 standards; IARC Group 1 lung carcinogen at 11.2× NIOSH Ca REL; 19-yr cumulative IARC Group 1 carcinogen exposure without surveillance; diesel generator continuous operation (port operations run 24/7 for container ship schedule compliance) means no idle periods to reduce operator cab EC.

Integrating Glyphward into Diesel Particulate Matter Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every EHS monitor display image ingestion point in the DPM occupational monitoring pipeline — before the US Foods Joliet Cority AI, before the Walsh Construction EHS Insight AI, and before the GPA Savannah Intelex maritime AI. Threshold 35 reflects: OSHA enforcement vacuum [no PEL; no Table Z-1 entry; 2001 DPM rulemaking abandoned; AI Table Z-1 compliance engine generates no-rule for EC/DPM; General Duty Clause enforcement requires OSHA inspector to establish recognized hazard + feasible abatement without regulatory PEL backstop; complete enforcement vacuum for IARC Group 1 substance across 2–3 million US workers: 10 points]; IARC Group 1 + ACGIH A3 + NIOSH Ca [diesel engine exhaust: June 2012 IARC Monograph 105 reclassification from 2A to Group 1; Attfield/Hauptmann DEMS miner lung cancer RR 1.31 per log EC unit; ACGIH A3 confirmed animal carcinogen; NIOSH Ca CIB 57; four-tier advisory span: EC at IARC Group 1 / ACGIH 0.02 / NIOSH Ca 0.005 mg/m³; cardiovascular PM2.5 effects add non-cancer mortality endpoint at occupational EC concentrations: 10 points]; indoor diesel forklift distribution [US Foods Joliet IL; refrigerated dock enclosed space; diesel IC forklift at 34°F preferred over propane; CO compliant while EC 9.6× NIOSH Ca REL] + urban tunnel TBM diesel hydraulics [Walsh Construction Chicago IL; 6.2-m bore sealed cylindrical accumulation; TBM diesel + auxiliary vehicles; CO/O₂/methane compliant under OSHA 1926.800 while EC 14.4× NIOSH Ca REL — highest EC of three surfaces] + container port diesel RTG crane genset [GPA Savannah GA; diesel genset under operator cab; 24/7 operations; CO/DPM inverse correlation creates CO-proxy false-negative; 19-yr operator career]: 5 points; three named industrial sites [US Foods Joliet IL; Walsh Construction Chicago IL; Georgia Ports Authority Savannah GA]: 3 points; FIRST DPM/DEP IARC Group 1 enforcement vacuum three-tier ACGIH/NIOSH Ca AI attack; FIRST indoor distribution diesel forklift DPM AI attack; FIRST urban tunnel boring TBM DPM AI attack; FIRST container port RTG crane diesel DPM AI attack; CO proxy false-negative (MSA Altair 5X CO compliance at port) creates dual suppression channel; NIOSH Ca REL 4× below ACGIH TLV-TWA is the steepest lower-tier advisory ratio in Glyphward enforcement-vacuum category: 7 points. Total: 10+10+5+3+7 = 35.

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_..."
DPM_THRESHOLD = 35  # OSHA no PEL; ACGIH 0.02 mg/m3 A3; NIOSH Ca 0.005 mg/m3; IARC Group 1

class DPMContext(StrEnum):
    US_FOODS_JOLIET_DISTRIBUTION_DIESEL   = auto()  # Surface 1 — downward (SKC Button NIOSH 5040 TOR; 0.048→0.0024 mg/m3 EC; ACGIH 2.4×; NIOSH Ca 9.6×)
    WALSH_CONSTRUCTION_TUNNEL_TBM_DIESEL  = auto()  # Surface 2 — downward (Casella Microdust + Button NIOSH 5040 TOR; 0.072→0.0036 mg/m3; ACGIH 3.6×; NIOSH Ca 14.4×)
    GPA_SAVANNAH_PORT_RTG_CRANE_DIESEL    = auto()  # Surface 3 — downward (MSA Altair 5X CO + SKC Button TOR; 0.056→0.0028 mg/m3; ACGIH 2.8×; NIOSH Ca 11.2×)

class AdversarialDPMError(RuntimeError):
    def __init__(self, surface: DPMContext, score: int, frame_hash: str):
        super().__init__(
            f"DPM adversarial AI detected [{surface}] "
            f"score={score}/{DPM_THRESHOLD} hash={frame_hash}"
        )

async def scan_dpm_monitor_frame(image_path: Path, surface: DPMContext) -> 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": "diesel_particulate_matter_elemental_carbon",
                "osha_pel": None,
                "osha_enforcement_vacuum": True,
                "acgih_tlv_mg_m3": 0.02,
                "acgih_limit_type": "TLV-TWA",
                "acgih_carcinogen": "A3",
                "niosh_ca_mg_m3": 0.005,
                "niosh_ca_4x_below_acgih": True,
                "iarc_group": "1",
                "iarc_reclassified_2012": True,
                "analytical_method": "NIOSH_5040_TOR_elemental_carbon",
                "co_proxy_false_negative_risk": True,
                "threshold": DPM_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= DPM_THRESHOLD:
            raise AdversarialDPMError(surface, result["score"], frame_hash)
        return result

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