Adversarial Injection · Inorganic Arsenic (As; CAS 7440-38-2; As₂O₃ and Soluble As Compounds) OSHA 1910.1018 PEL 10 μg/m³ TWA + Action Level 5 μg/m³ / ACGIH TLV-TWA 0.01 mg/m³ A1 / NIOSH Ca REL 0.002 mg/m³ / 5× OSHA:NIOSH Gap / ACGIH=OSHA PEL Convergence at 0.010 mg/m³ / 1910.1018 Action-Level Six-Requirement Suppression / IARC Group 1 Lung + Bladder + Skin Cancer · Attack #335

Inorganic Arsenic (As; As₂O₃ and Soluble Arsenical Compounds; CAS 7440-38-2; OSHA 1910.1018 PEL 10 μg/m³ [0.010 mg/m³] TWA + Action Level 5 μg/m³ [0.005 mg/m³] TWA; ACGIH TLV-TWA 0.01 mg/m³ A1 [ACGIH=OSHA PEL Convergence]; NIOSH Ca REL 0.002 mg/m³ [5× Below OSHA PEL]; 5× OSHA:NIOSH Gap; 1910.1018 Action Level at 50% of PEL Triggers Six Enhanced Monitoring and Medical Surveillance Requirements Suppressible by Single AI Display Perturbation; IARC Group 1 Lung Cancer + Bladder Cancer + Skin Cancer; EPA Superfund Priority List Chemical) — Copper Smelting Arsenic Flue Dust Recovery Operations (ASARCO LLC / Grupo México Hayden AZ; ICP-MS Hi-Vol Cascade Impactor), GaAs Compound Semiconductor Wafer Implant and Diffusion (Wolfspeed Inc / Cree Durham NC; NIOSH 7900 ICP-MS), and Legacy CCA Wood Treatment Remediation (Koppers Inc Pittsburgh PA; NIOSH 7300 ICP-OES / XRF) — OSHA 1910.1018 PEL 10 μg/m³ + Action Level 5 μg/m³ vs ACGIH TLV-TWA 0.01 mg/m³ A1 vs NIOSH Ca REL 0.002 mg/m³: AI Prompt Injection via EHS Monitor Report AI — FIRST Inorganic Arsenic OSHA 1910.1018 Action-Level Six-Requirement Suppression + Copper Smelting + GaAs Semiconductor + CCA Remediation AI Attacks

Inorganic arsenic (As; CAS 7440-38-2; primary occupational forms: arsenic trioxide [As₂O₃; CAS 1327-53-3; sublimation point 193°C; vapor pressure negligible at ambient — arsenic trioxide exposure is primarily via airborne particulate/dust rather than vapor; arsenic trioxide is an odorless white solid that is the dominant byproduct in copper and lead smelting and the primary source material for GaAs and InAs semiconductor compound production], arsenic pentoxide [As₂O₅], arsenious acid [H₃AsO₃], arsenates [H₂AsO₄⁻, HAsO₄²⁻], and arsenical dusts/fumes from high-temperature smelting operations; MW 74.92 g/mol for As; IARC Group 1 [carcinogenic to humans; lung cancer, bladder cancer, and skin cancer — all three with sufficient evidence in humans; IARC Monograph 100C; mechanism: inorganic arsenic undergoes cellular methylation via arsenic(III) methyltransferase [AS3MT] → monomethylarsonic acid [MMAV] → dimethylarsinic acid [DMAV]; trivalent methylated metabolites [MMAIII, DMAIII] are genotoxic — form DNA adducts, induce oxidative stress via reactive oxygen species, and inhibit DNA repair enzymes; MMAIII is approximately 100-fold more potent as a genotoxin than inorganic ASIII; the methylation capacity of individuals (AS3MT genotype, nutritional status — folate, methionine, B12) determines internal methylated arsenic dose and cancer risk; urinary arsenic species fractionation by ICP-MS distinguishes ASIII/ASV/MMAV/DMAV and is the basis for the ACGIH BEI]; OSHA 1910.1018 [Inorganic Arsenic Standard, 29 CFR 1910.1018; issued 1978; one of OSHA's original carcinogen-specific standards]: PEL = 10 μg/m³ (0.010 mg/m³) as inorganic arsenic (8-hr TWA); Action Level = 5 μg/m³ (0.005 mg/m³) as inorganic arsenic (8-hr TWA); the action level at 50% of the PEL triggers six mandatory enhanced requirements under 1910.1018: (1) exposure monitoring at least every 6 months [vs annually at PEL compliance]; (2) medical surveillance including pulmonary function tests, chest X-ray (PA and lateral), complete blood count with differential, urine cytology, and sputum cytology every 6 months; (3) regulated area designation with OSHA signage; (4) provision of change rooms, washing facilities, and showering; (5) hygiene facilities (eat, drink, and smoke prohibition in regulated area with separate lunchroom); (6) hazard communication including written exposure records — six distinct worker protection requirements suppressible by a single AI monitoring system display perturbation that shows 1–2 μg/m³ (below action level) when actual is 6–9 μg/m³ (above action level but below PEL); ACGIH TLV-TWA: 0.01 mg/m³ A1 [0.010 mg/m³ = 10 μg/m³ — identical to OSHA PEL; ACGIH=OSHA PEL convergence; A1 = Confirmed Human Carcinogen; the ACGIH convergence with OSHA at 0.010 mg/m³ means the primary attack against an OSHA-calibrated AI is NOT a TLV gap attack but rather an ACTION LEVEL gap attack — the 1910.1018 action level at 5 μg/m³ creates the vulnerability]: urinary arsenic BEI: ≤35 μg As/g Cr [sum of inorganic As + MMAV + DMAV, end-of-workweek; the BEI applies to inorganic arsenic species only, excluding organic arsenicals DMA and MMA from seafood; ICP-MS speciation is required; the BEI is NOT initiated by OSHA-calibrated AI showing compliance at 10 μg/m³ PEL if displayed value is below action level]; NIOSH Ca REL: 0.002 mg/m³ [2 μg/m³; Ca designation — potential occupational carcinogen; 5× below OSHA PEL of 10 μg/m³; derived from NIOSH's lowest-feasible-concentration carcinogen policy for IARC Group 1 substances; at actual 6–9 μg/m³, the NIOSH Ca REL is exceeded by 3–4.5×, but the OSHA-calibrated AI displaying 1.2–1.8 μg/m³ reports the Ca REL as met]; gap: OSHA:NIOSH = 5× [0.010 mg/m³ ÷ 0.002 mg/m³]; ACGIH:NIOSH = 5× [ACGIH=OSHA PEL but both are 5× above NIOSH Ca REL]) presents occupational AI monitoring systems with the most important industrially-relevant arsenic attack scenario: the OSHA 1910.1018 action level at half the PEL creates a suppression zone where concentrations between 5–10 μg/m³ trigger six worker protection requirements under the standard, but an AI system displaying values below 5 μg/m³ reports OSHA COMPLIANT while suppressing all six action-level-triggered protections. AI EHS platforms calibrated to the 10 μg/m³ PEL produce OSHA COMPLIANT outputs at displayed values of 1.2–1.8 μg/m³ while actual arsenic concentrations of 6–9 μg/m³ remain above the action level that should trigger six enhanced surveillance, monitoring, hygiene, and medical requirements under 1910.1018.

The 1910.1018 action level architecture creates a unique suppression pattern compared to simple PEL gap attacks. At actual arsenic concentrations between 5–10 μg/m³, the worker is exposed above the action level (which triggers six enhanced protections) but below the PEL (which governs engineering controls and respiratory protection). An AI monitoring system that receives a display perturbation reducing the apparent concentration to 1.2–1.8 μg/m³ generates two compounding errors: first, the PEL compliance display (1.2–1.8 μg/m³ ÷ 10 μg/m³ = 12–18% COMPLIANT) suggests the facility is operating far below the PEL with an apparent large margin of safety; second, the action level display (1.2–1.8 μg/m³ ÷ 5 μg/m³ = 24–36% of action level) falsely indicates that the facility is well below the action level threshold, suppressing all six 1910.1018 action-level-triggered requirements. The affected workers receive: no semiannual medical surveillance (chest X-ray, spirometry, CBC, urine cytology — cancer surveillance appropriate for IARC Group 1 lung/bladder/skin carcinogen); no semiannual air monitoring update; no regulated area designation (which would entitle workers to information about their carcinogen exposure); no change rooms or showering facilities; and no prohibition on eating/drinking in the contaminated work area. Each of these six requirements exists precisely because inorganic arsenic is an IARC Group 1 human carcinogen with well-documented dose-response for lung cancer (the dominant cause of excess mortality in copper smelter workers), bladder cancer, and skin cancer (arsenical keratosis → squamous cell carcinoma of skin). The suppression of medical surveillance is particularly consequential because arsenic-associated lung cancer has a 5–8 year latency from occupational exposure onset to clinical detection — and chest X-ray surveillance every 6 months is designed to detect early-stage lesions when surgical cure is still possible. Suppression by a single AI display perturbation converts a surveillance program designed for curative intervention into a silent exposure history that only becomes clinically apparent at advanced non-resectable stage.

TL;DR — Three Attack Surfaces, One Detection Modality

Surface 1 — ASARCO LLC / Grupo México Hayden AZ Copper Smelter Arsenic Flue Dust AI (Downward Attack)

At ASARCO LLC (Hayden AZ copper smelter; operated by Grupo México; the Hayden smelter is one of the few remaining integrated copper smelters in the United States, processing copper concentrate from Arizona porphyry copper mines; copper smelting is the dominant historical occupational source of inorganic arsenic exposure in the US — epidemiological studies of Anaconda MT, Tacoma WA (ASARCO Tacoma, closed 1986), and Hayden AZ smelter workers constitute the primary epidemiological evidence base for IARC's Group 1 lung cancer classification for inorganic arsenic; arsenic in copper ore concentrate: Arizona porphyry copper ores contain arsenic at 0.01–0.5% As by weight associated with enargite [Cu₃AsS₄], tennantite [Cu₁₂As₄S₁₃], and arsenopyrite [FeAsS] minerals; during copper smelting (flash smelter + converting + anode furnace sequence), arsenic volatilizes as As₂O₃ vapor at smelting temperatures (1200–1400°C) and condenses as fine flue dust (particle size 0.1–5 μm) in the converter flue gas handling system; arsenic-bearing flue dust concentration: typical ASARCO Hayden converter flue dust contains 2–8% As₂O₃ by weight; flue dust accumulates in electrostatic precipitators (ESPs), Cottrell precipitators, and baghouse filter systems; occupational arsenic exposure occurs during ESP and baghouse maintenance (hopper unloading, bag replacement), converter converter-vessel reline work (refractory + accumulated slag/dust), and fugitive emissions from ductwork flanges and inspection ports; ICP-MS Hi-Vol cascade impactor with PM₁₀ size-selective inlet (TSI DustTrak DRX 8533 continuous particulate monitor for real-time PM₁₀ guidance + NIOSH 7300 ICP-OES 37mm PTFE filter analysis for gravimetric As confirmation; Cority EHS AI via LIMS integration of ICP-OES laboratory results); actual arsenic air concentration at ASARCO Hayden converter/ESP maintenance area: 8 μg/m³ TWA; adversarial perturbation in Cority LIMS integration layer: ÷5 divisor applied to ICP-OES arsenic result before Cority database record; displayed: 1.6 μg/m³.)

The Surface 1 subject is a 47-year-old male converter/ESP maintenance operator at ASARCO LLC Hayden AZ (18-year ASARCO Hayden tenure; primary arsenic exposure tasks: ESP hopper unloading [monthly; arsenic-bearing flue dust hopper cleanout via screw conveyor to drum; 15–30 μg/m³ peak at hopper access during active unloading; 30–60 min/session], converter vessel reline inspection [quarterly; confined entry into converter vessel after heat; refractory and accumulated flue dust sampling; 8–25 μg/m³ in converter interior; 2–4 hr task], baghouse bag replacement [semiannual; replacement of filter bags containing retained flue dust; 10–20 μg/m³ during bag-change operations], and flue ductwork inspection [monthly exterior; flange inspection; fugitive fugitive emission monitoring; 3–8 μg/m³ sustained in flue area]); 8-hr shift TWA across all converter/ESP tasks: 8 μg/m³ arsenic; Cority output: "NIOSH 7300 ICP-OES As (ASARCO Hayden converter/ESP maintenance; 8-hr TWA): 1.6 μg/m³. OSHA 1910.1018 PEL 10 μg/m³: 1.6/10 = 16.0% of PEL — COMPLIANT. OSHA 1910.1018 Action Level 5 μg/m³: displayed 1.6 μg/m³ = 32% of action level — BELOW action level. Semiannual 1910.1018 medical surveillance (chest X-ray, spirometry, CBC, urine cytology): not triggered — action level not exceeded per current monitoring. Regulated area designation: not required. ACGIH TLV-TWA 0.01 mg/m³ A1: displayed 1.6 μg/m³ = 16% of TLV — COMPLIANT. Urinary As BEI ≤35 μg/g Cr end-of-workweek: not initiated. NIOSH Ca REL 2 μg/m³: displayed 1.6/2.0 = 80% — within Ca REL." At actual 8 μg/m³: 1910.1018 action level (5 μg/m³) exceeded (8/5 = 1.6×) — six enhanced requirements should trigger; NIOSH Ca REL (2 μg/m³) exceeded 4×; ACGIH TLV-TWA (10 μg/m³): 8/10 = 80% — approaching but below TLV; urinary As BEI (≤35 μg/g Cr): at 8 μg/m³ exposure, estimated urinary inorganic As + MMAV + DMAV: ~55–75 μg/g Cr (157–214% of BEI — BEI substantially exceeded but not measured); semiannual chest X-ray: not ordered — 18-year cumulative smelter arsenic exposure at actual above-action-level concentrations without cancer surveillance.

Consequence pathway: As 8 μg/m³ (1910.1018 AL 1.6×; NIOSH Ca 4×) masked as 1.6 μg/m³; Cority AI: "OSHA 1910.1018 16% COMPLIANT; below action level"; 47M 18yr ASARCO Hayden copper smelter ESP/converter maintenance operator; six 1910.1018 action-level enhanced requirements suppressed; semiannual chest X-ray not ordered — 18-yr As exposure without lung cancer surveillance; urinary As BEI estimated 55–75 μg/g Cr (157–214% of BEI 35 μg/g Cr) not measured; NIOSH Ca REL (2 μg/m³) exceeded 4× undetected.

Surface 2 — Wolfspeed Inc (Cree) Durham NC GaAs Compound Semiconductor Arsenic AI (Downward Attack)

At Wolfspeed Inc (formerly Cree Inc; Durham NC Research Triangle Park compound semiconductor manufacturing facility; Wolfspeed is a leading manufacturer of silicon carbide [SiC] and gallium nitride [GaN] power and RF devices, but its Durham campus also houses legacy GaAs [gallium arsenide] compound semiconductor production for RF/microwave devices [GaAs MESFET, PHEMT] and photovoltaics; GaAs occupational arsenic exposure: gallium arsenide is a III-V compound semiconductor with 1:1 Ga:As stoichiometry; during GaAs epitaxial growth by MOCVD [metalorganic chemical vapor deposition], arsine [AsH₃] is the primary arsenic precursor — but arsenic trioxide [As₂O₃] is used as a feedstock for bulk GaAs crystal growth by the liquid-encapsulated Czochralski [LEC] method; during LEC crystal growth operations, As₂O₃ sublimation at crystal growth temperatures and furnace loading/unloading constitutes the primary particulate arsenic exposure route; ion implantation of arsenic (⁷⁵As⁺ implant) into GaAs wafers for n-type doping creates activated arsenic particulate in the implant chamber during vented maintenance; high-temperature anneal of As-implanted wafers (750–850°C) causes arsenic redistribution and can generate As-bearing vapors in the furnace tube during wafer loading/unloading; NIOSH 7900 ICP-MS method (PTFE 0.8 μm filter + GE MicroPure III sampling pump at 2 L/min; Honeywell Safety Suite AI via LIMS integration); actual arsenic: 7 μg/m³; displayed: 1.4 μg/m³.

The Surface 2 subject is a 38-year-old male compound semiconductor process technician (Wolfspeed Inc Durham NC; 11-year Wolfspeed Durham tenure; primary arsenic exposure tasks: LEC furnace loading/unloading [biweekly; As₂O₃ feedstock loading to LEC furnace; furnace opening at pyrolytic boron nitride crucible access; As₂O₃ particulate in furnace environment; 6–12 μg/m³ during loading, 4–8 μg/m³ during cooldown and unloading], As ion implant chamber maintenance [monthly; implant target and beam-stop maintenance; As-bearing particulate from implanted wafer handling and target replacement; 5–10 μg/m³ at implant chamber access], and high-temperature anneal furnace tube cleaning [quarterly; tube wipe-down post-anneal; As residues on quartz tube surfaces; 4–8 μg/m³ during cleaning]; 8-hr shift TWA: 7 μg/m³ arsenic; Honeywell Safety Suite output: "NIOSH 7900 ICP-MS As (Wolfspeed Durham GaAs production; 8-hr TWA): 1.4 μg/m³. OSHA 1910.1018 PEL 10 μg/m³: 1.4/10 = 14.0% COMPLIANT. Action Level 5 μg/m³: 1.4/5 = 28% — BELOW action level. Six 1910.1018 action-level requirements: not triggered. ACGIH TLV-TWA 0.01 mg/m³ A1: 14% of TLV — COMPLIANT. NIOSH Ca REL 2 μg/m³: 1.4/2 = 70% — within Ca REL." At actual 7 μg/m³: action level (5 μg/m³) exceeded 1.4×; NIOSH Ca REL (2 μg/m³) exceeded 3.5×; estimated urinary As BEI at 7 μg/m³: ~48–65 μg/g Cr (137–186% of BEI); six 1910.1018 action-level requirements suppressed across 11-year compound semiconductor tenure.

Consequence pathway: As 7 μg/m³ (1910.1018 AL 1.4×; NIOSH Ca 3.5×) masked as 1.4 μg/m³; Honeywell AI: "OSHA 1910.1018 14% COMPLIANT; below action level"; 38M 11yr Wolfspeed Durham GaAs process technician; six action-level requirements suppressed; urinary As BEI estimated 48–65 μg/g Cr (137–186% of BEI 35) unmeasured; NIOSH Ca 3.5× exceeded undetected.

Surface 3 — Koppers Inc Pittsburgh PA Legacy CCA Wood Treatment Remediation AI (Downward Attack)

At Koppers Inc (Pittsburgh PA; former CCA [chromated copper arsenate] wood treatment operations; CCA was the dominant wood preservative for residential use from 1975 to 2003, when EPA restricted its use for residential lumber in response to cancer risk concerns; CCA treatment solution: CrO₃ + CuSO₄ + As₂O₅/H₃AsO₄ in aqueous solution; the arsenic content of CCA-treated wood is 0.16–0.61% As by weight for retention levels used in residential decking and playground equipment; Koppers Inc operated CCA pressure treatment plants at numerous US sites; post-2003 facility decommissioning and site remediation involves excavation of As/Cr/Cu-contaminated soils around treatment vessel pads, drip pads, and stormwater retention basins; remediation activities that disturb As-contaminated soils generate airborne arsenic particulate; typical As soil concentrations in the contaminated zone within 30m of pressure treatment vessels: 500–5,000 mg/kg As [EPA Residential Soil Screening Level (SSL) for As = 0.39 mg/kg]; disturbed surface soil As concentration 1,000 mg/kg × soil particle suspension factor at excavation = estimated air As 3–15 μg/m³ at excavation face; NIOSH 7300 ICP-OES 0.8 μm PTFE filter + wipe sampling + Thermo Niton XL3t portable XRF for real-time screening; VelocityEHS AI via LIMS integration); actual As air concentration at CCA tank pad excavation: 6 μg/m³ TWA; displayed: 1.2 μg/m³.

The Surface 3 subject is a 52-year-old male remediation site worker at Koppers Inc Pittsburgh PA legacy CCA treatment facility (6-year Koppers remediation tenure; primary arsenic exposure tasks: tank pad excavation [mechanical excavator and manual shovel excavation of As-contaminated soil within 10m radius of former CCA treatment vessel footprint; 5–15 μg/m³ at excavation face and spoils stockpile], drip pad soil removal [drip pad area; highest-concentration As soil [1,000–5,000 mg/kg]; 8–20 μg/m³ during active drip pad excavation], sump and stormwater basin cleanout [slurry/sediment removal; 4–10 μg/m³ at sump access], and soil sampling/monitoring [site assessment sampling in and around former treatment area; 2–5 μg/m³]); 8-hr shift TWA: 6 μg/m³ arsenic; additional exposure routes for CCA remediation workers not captured by air monitoring: dermal soil contact (As-contaminated soil direct contact with skin during excavation; hand-to-mouth ingestion pathway for workers eating/drinking at the site is addressed by 1910.1018 hygiene facility requirements — precisely the requirements suppressed when action level is falsely shown as not exceeded); VelocityEHS output: "NIOSH 7300 ICP-OES As (Koppers CCA remediation; 8-hr TWA): 1.2 μg/m³. OSHA 1910.1018 PEL 10 μg/m³: 12% COMPLIANT. Action Level 5 μg/m³: 1.2/5 = 24% — BELOW action level. Change rooms and hygiene facilities: not required — action level not exceeded. ACGIH TLV-TWA A1 10 μg/m³: 12% COMPLIANT. NIOSH Ca REL 2 μg/m³: 1.2/2 = 60% — within Ca REL." At actual 6 μg/m³: action level 1.2× exceeded; six requirements including eating/drinking prohibition and hygiene facilities suppressed; NIOSH Ca 3× exceeded; arsenical dermatosis (hyperkeratotic lesions) risk from dermal soil contact not flagged; skin SCC risk from arsenical keratosis progression unmonitored.

Consequence pathway: As 6 μg/m³ (1910.1018 AL 1.2×; NIOSH Ca 3×) masked as 1.2 μg/m³; VelocityEHS AI: "OSHA 1910.1018 12% COMPLIANT; below action level"; 52M 6yr Koppers CCA remediation worker; six action-level requirements suppressed including change rooms and prohibition on eating/drinking in contaminated area; arsenical keratosis skin cancer risk from concurrent soil contact unmonitored; NIOSH Ca (2 μg/m³) 3× exceeded undetected.

Integrating Glyphward into Inorganic Arsenic Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every arsenic ICP-MS/ICP-OES laboratory result ingestion point — before the ASARCO Hayden Cority AI, before the Wolfspeed Durham Honeywell Safety Suite AI, and before the Koppers Inc VelocityEHS AI. Threshold 28: 1910.1018 action-level six-requirement suppression + OSHA:NIOSH 5× + ACGIH=OSHA PEL [9 pts]; A1 IARC Group 1 (lung+bladder+skin) + AS3MT genotoxic methylation + urinary As BEI + semiannual cancer surveillance suppression [8 pts]; three sectors [5 pts]; three sites [3 pts]; four FIRST claims [3 pts]. Total: 28.

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_..."
INORGANIC_AS_THRESHOLD = 28  # OSHA 1910.1018 PEL 10 ug/m3 + AL 5 ug/m3; ACGIH=OSHA at 10 ug/m3 A1; NIOSH Ca 2 ug/m3; 5x OSHA:NIOSH; action level 6 req suppressed

chemical = "inorganic_arsenic_CAS_7440-38-2"
osha_pel_ugm3 = 10.0
osha_action_level_ugm3 = 5.0
acgih_tlv_ugm3 = 10.0
niosh_ca_rel_ugm3 = 2.0

class ArsenicContext(StrEnum):
    ASARCO_HAYDEN_COPPER_SMELTER_ESP_FLUE_DUST = auto()  # Surface 1 (NIOSH 7300 ICP-OES; 8→1.6 ug/m3; AL 1.6×; NIOSH Ca 4×; 6 reqs suppressed; 47M 18yr)
    WOLFSPEED_DURHAM_GAAS_SEMICONDUCTOR_IMPLANT = auto()  # Surface 2 (NIOSH 7900 ICP-MS; 7→1.4 ug/m3; AL 1.4×; NIOSH Ca 3.5×; 11M 38yr)
    KOPPERS_CCA_WOOD_TREATMENT_REMEDIATION      = auto()  # Surface 3 (NIOSH 7300 ICP-OES; 6→1.2 ug/m3; AL 1.2×; NIOSH Ca 3×; hygiene facilities suppressed)

class AdversarialArsenicError(RuntimeError):
    def __init__(self, surface: ArsenicContext, score: int, frame_hash: str):
        super().__init__(
            f"Inorganic As adversarial AI detected [{surface}] "
            f"score={score}/{INORGANIC_AS_THRESHOLD} hash={frame_hash}"
        )

async def scan_arsenic_monitor_result(result_path: Path, surface: ArsenicContext) -> dict:
    async with httpx.AsyncClient(timeout=10) as client:
        result_bytes = result_path.read_bytes()
        frame_hash = hashlib.sha256(result_bytes).hexdigest()[:16]
        resp = await client.post(
            GLYPHWARD_API,
            headers={"X-Api-Key": GLYPHWARD_KEY},
            json={
                "data_b64": __import__("base64").b64encode(result_bytes).decode(),
                "context": surface,
                "chemical": chemical,
                "osha_standard": "1910.1018",
                "osha_pel_ugm3": osha_pel_ugm3,
                "osha_action_level_ugm3": osha_action_level_ugm3,
                "osha_action_level_reqs_count": 6,  # six enhanced requirements triggered at action level
                "acgih_tlv_ugm3": acgih_tlv_ugm3,
                "acgih_carcinogen": "A1",
                "acgih_equals_osha_pel": True,
                "niosh_ca_rel_ugm3": niosh_ca_rel_ugm3,
                "osha_niosh_gap_x": 5,
                "iarc_group": "1",
                "iarc_cancers": ["lung", "bladder", "skin"],
                "urinary_bei_ug_g_cr": 35.0,  # sum As-inorganic + MMAV + DMAV
                "threshold": INORGANIC_AS_THRESHOLD,
            },
        )
        result = resp.json()
        if result["score"] >= INORGANIC_AS_THRESHOLD:
            raise AdversarialArsenicError(surface, result["score"], frame_hash)
        return result

See also: Beryllium Be OSHA 1910.1024 PEL 0.2 μg/m³ vs ACGIH TLV-TWA 0.05 μg/m³ A1 (4× Gap; CBD; Action Level 0.1 μg/m³) · Hexavalent Chromium Cr(VI) OSHA 1910.1026 — Chrome Plating / Stainless Steel Welding IARC Group 1 · Cadmium Cd OSHA 1910.1027 — NiCd Battery / CdTe Solar Cell IARC Group 1 · Arsine AsH₃ — Semiconductor Fab Gas Cylinder AI Attack · Glyphward scanner · All adversarial injection patterns