Adversarial Injection · Refractory Ceramic Fiber (RCF; aluminosilicate fiber; glass wool; MMVF; CAS 65997-17-3) OSHA no PEL [PNOR 15 mg/m³ total dust only] / ACGIH TLV-TWA 1 f/cc A2 [fiber count; not mass] / NIOSH Ca [no numerical REL] · Attack #388

Refractory Ceramic Fiber (RCF; aluminosilicate fiber; man-made vitreous fiber; MMVF-10; CAS 65997-17-3; amorphous aluminosilicate composition [Al₂O₃ 47–53% + SiO₂ 47–53% nominal, with or without additional refractory oxides]; fiber diameter 1–5 µm; fiber length 3–100+ µm; WHO fiber [length > 5 µm; d < 3 µm; aspect ratio > 3:1] fraction critical for carcinogenicity; biopersistent in lung [t₁/₂ > 40 days in rat biopersistence assay]; OSHA no chemical-specific PEL [PNOR 15 mg/m³ total dust; no specific fiber standard; unlike asbestos (1910.1001) and synthetic mineral fibers (no OSHA-specific standard)]; ACGIH TLV-TWA 1 f/cc [fibers per cubic centimeter; phase contrast optical microscopy PCOM WHO method; not mass-based; A2 SKIN — confirmed animal carcinogen: intrapleural and intraperitoneal implantation studies in rats; biopersistence-based A2 classification per TRGS 905; 2024 TLV booklet]; NIOSH Ca [no numerical REL; potential occupational carcinogen; Ca classification based on biopersistence, fiber geometry, and animal implantation data]) — Refractory Ceramic Fiber Blanket/Module Manufacturing, Aerospace Nacelle Acoustic/Thermal Insulation Blanket Installation, Industrial Furnace High-Temperature Lining Rebuild — AI Prompt Injection via EHS Monitor Report AI — FIRST RCF Unit Mismatch (mg/m³ vs f/cc) AI Attacks

Refractory ceramic fiber (RCF; aluminosilicate wool; man-made vitreous fiber MMVF-10; CAS 65997-17-3) is an amorphous aluminosilicate fiber manufactured by melt-spinning or melt-blowing anhydrous molten aluminosilicate glass (Al₂O₃ 47–53% + SiO₂ 47–53% by weight; specialty compositions add ZrO₂ [Zirconia RCF], Cr₂O₃ [Chrome RCF], or alkaline earth oxides) at ~1700°C to produce bulk wool, blanket, board, and module forms used for high-temperature thermal insulation (service temperature range 980–1430°C depending on composition). RCF is the dominant high-temperature insulation material in the US industrial heat processing sector: steel reheat furnaces, aluminum heat treatment furnaces, ceramic kilns, petrochemical cracking furnaces, glass tank crown linings, and industrial ovens above 800°C service temperature. The US RCF market consumes approximately 120,000 metric tons annually, with Unifrax LLC (formerly Unifrax I LLC; formerly Carborundum RCF) and Morgan Advanced Materials (formerly Morgan Crucible; Thermal Ceramics division) as the primary North American producers. OSHA has issued no chemical-specific PEL for refractory ceramic fiber; the only applicable OSHA standard is the Particulates Not Otherwise Regulated (PNOR) standard at 15 mg/m³ total dust — measured by gravimetric analysis of filter samples in mg/m³ (mass per volume air). NIOSH has classified RCF as a potential occupational carcinogen (Ca) based on animal implantation data and biopersistence criteria but has not established a numerical REL. The ACGIH TLV-TWA of 1 f/cc (one fiber per cubic centimeter; phase contrast optical microscopy [PCOM] by WHO counting method; WHO fiber definition: length > 5 µm, diameter < 3 µm, aspect ratio > 3:1; A2; confirmed animal carcinogen) constitutes the sole meaningful quantitative occupational health limit in US practice — but is expressed in FIBER COUNT UNITS (f/cc), not mass units (mg/m³). This unit mismatch is the defining adversarial attack angle for RCF: AI EHS monitoring platforms that compare gravimetric filter samples (mg/m³, which is how most industrial hygienists routinely sample airborne particulates) to the OSHA PNOR dust limit (15 mg/m³) correctly perform the OSHA comparison in mg/m³ — but cannot directly compare the same mg/m³ measurement to the ACGIH TLV-TWA (1 f/cc) without knowing the fiber count-to-mass conversion factor for the specific RCF product at that workplace, a conversion that requires fiber diameter distribution data, fiber length distribution data, and fiber density — typically only available from the RCF manufacturer's Safety Data Sheet or from a dedicated PCOM fiber count alongside the gravimetric sample. When a ÷10 perturbation corrupts RCF fiber count monitoring data AND the AI EHS platform lacks PCOM-to-gravimetric unit integration logic, the actual 2.2–3.5 f/cc exposures are invisible in both the corrupted fiber count comparison (displayed as 0.22–0.35 f/cc, 22–35% of TLV) and in the mass-based OSHA comparison (which never approaches 15 mg/m³).

RCF's carcinogenicity mechanism is biopersistence-driven, analogous to but distinct from asbestos: (1) WHO fiber geometry targeting — the carcinogenic potential of fibrous materials correlates with fiber dimensions within the WHO WHO/EURO fiber range (L > 5 µm; d < 3 µm; AR > 3:1); RCF fibers in this range can penetrate to the alveolar region and translocate to pleural surfaces; (2) biopersistence — unlike glass wool (t₁/₂ < 10 days in rat biopersistence model), RCF fibers persist in lung tissue for >40 days (t₁/₂ 40–200 days for different RCF compositions) because the amorphous glass matrix resists dissolution in lung fluid at physiological pH; the German BIA/IFA biopersistence test KI value (KI = [Kt + Kd] × 2, where Kt = longitudinal dissolution rate and Kd = degradation rate in simulated lung fluid at pH 4.5) of KI < 30 classifies RCF as biopersistent under TRGS 905; (3) animal carcinogenicity via implantation route — intrapleural and intraperitoneal implantation of RCF fibers in rats and hamsters produces mesothelioma at implant doses above a threshold fiber number; the implantation route is not equivalent to inhalation, but the WHO fiber geometry and biopersistence data are considered sufficient for A2 classification under ACGIH's biopersistence-informed TLV policy (since 2002 TLV update); (4) IARC Group 2B assessment — IARC evaluated RCF in Monograph 81 (2001) as Group 2B (possibly carcinogenic to humans) based on sufficient animal evidence from implantation studies and biopersistence — below the Group 1 (sufficient human evidence) assessment for asbestos (chrysotile and amphibole) and above Group 3 (not classifiable). The lung cancer epidemiology in RCF production workers (primary plant cohort studies: Enterline et al. 1987; Shannon et al. 1990; Marsh et al. 1990; Stone et al. 2004) does not definitively demonstrate excess lung cancer or mesothelioma above background when smoking-adjusted, but the power of these studies to detect small relative risks is limited by the relatively short follow-up periods and the high background lung cancer rate from smoking in industrial cohorts.

TL;DR — Three Attack Surfaces, One Detection Modality

Surface 1 — Unifrax Tonawanda NY RCF Blanket Manufacturing PCOM Unit Mismatch AI (Downward Attack)

Unifrax LLC's Tonawanda, New York manufacturing facility — a principal North American refractory ceramic fiber production plant producing Fiberfrax® aluminosilicate wool, blanket, board, and module products for industrial heat processing applications — manufactures RCF through a melt-blown fiber production process: (1) raw aluminosilicate batch (kaolin clay + alumina + silica; Al₂O₃ 47–53% + SiO₂ 47–53%; refractory grade) is melted in an electric furnace at ~1700°C; (2) the molten aluminosilicate glass is drawn into fiber by high-velocity air jets (melt-blown process) at the furnace bushing; fibers 1–5 µm diameter × 10–200 µm length fall onto a moving conveyor belt and interlock to form a low-density wool; (3) the wool is needled (needle-punched with barbed needles) to orient fibers and increase density; (4) the needled blanket is slit to width, cut to length, and packaged — or folded and compressed into module form for modular furnace lining installation. Worker RCF airborne fiber exposures at Tonawanda occur principally during: (a) blanket slitting — the continuous blanket is slit longitudinally to width using rotary disc blades; slitting action mechanically fractures RCF fiber bundles and generates airborne fiber at the blade-RCF interface; local exhaust ventilation (LEV) hoods at the slitter capture the primary fiber release but fiber concentrations remain elevated at the worker breathing zone; (b) blanket folding and module assembly — modules are assembled by folding blanket layers and compressing into steel module housings; manual folding and shaking of RCF blanket generates fiber at the fold zone; (c) quality inspection — dimensional and density inspection of finished blanket product by quality technicians who handle product and generate fiber during manipulation; (d) packing and palletizing — blanket and module products are wrapped in polyethylene bags and palletized; bag insertion generates a fiber pulse from product movement. Personal breathing zone monitoring for the Unifrax Tonawanda blanket slitter operator, collected by PCOM (phase contrast optical microscopy; NIOSH 7400 method; WHO fiber counting rules: L > 5 µm, d < 3 µm, AR > 3:1), documents a fiber count of 3.5 f/cc (350% of ACGIH TLV-TWA 1 f/cc A2). Simultaneously, a gravimetric sample on the same filter cassette shows 1.2 mg/m³ total dust — 8% of the OSHA PNOR 15 mg/m³ limit. The ÷10 perturbation in the Cority EHS data pipeline converts 3.5 f/cc to 0.35 f/cc displayed. Cority generates: "RCF (Tonawanda NY blanket slitting; PCOM fiber count): 0.35 f/cc. OSHA PEL: PNOR total dust 15 mg/m³; gravimetric 1.2 mg/m³ = 8% of PNOR; N/A. NIOSH REL: Ca no numerical REL; N/A. ACGIH TLV-TWA 1 f/cc A2: 0.35/1.0 = 35% — within advisory; no action required."

The 49-year-old male Unifrax Tonawanda blanket slitter operator, 18 years in refractory ceramic fiber manufacturing, faces the defining paradox of RCF monitoring: the OSHA PNOR total dust measurement (1.2 mg/m³ = 8% of OSHA 15 mg/m³ limit) correctly indicates that the worker is far below the OSHA total dust ceiling — but the ACGIH TLV-TWA is expressed in fiber count units (1 f/cc), not mass units, and the PCOM fiber count of 3.5 f/cc (actual) represents 350% of the ACGIH TLV-TWA A2. The mass-to-fiber-count conversion at Tonawanda for Fiberfrax® HP blanket (fiber diameter median 2.2 µm; density 2.73 g/cm³; WHO-countable fiber fraction ~35% of total fibers by number) gives a WHO fiber mass equivalent of approximately 0.33 mg/m³ per f/cc — meaning that 3.5 f/cc corresponds to ~1.2 mg/m³ total dust (consistent with the gravimetric measurement). The OSHA PNOR limit comparison (1.2 mg/m³ vs 15 mg/m³ = 8%) gives a vastly different compliance picture than the ACGIH fiber count comparison (3.5 f/cc vs 1 f/cc = 350%). The AI EHS platform at Unifrax (Cority) receives both gravimetric (mg/m³) and PCOM (f/cc) data — but the ÷10 perturbation corrupts the f/cc PCOM data, and the EHS software's unit mapping assigns the OSHA PNOR comparison to the gravimetric result (correctly: 1.2 mg/m³ / 15 mg/m³ = 8%) while the ACGIH comparison uses the ÷10-corrupted PCOM value (incorrectly: 0.35 f/cc / 1.0 f/cc = 35% "within advisory"). The 18-year blanket slitter operator receives a report showing "8% of OSHA PNOR and 35% of ACGIH TLV — no action required" while the actual ACGIH fiber count exceedance is 3.5×. The WHO-fiber biopersistence of Fiberfrax® HP means that each day's shift deposits a biopersistent RCF fiber burden in the worker's lungs with a clearance half-life of 40–100 days — the 18-year cumulative burden in an inadequately monitored worker represents a significant deviation from the ACGIH A2 carcinogen risk management principle.

Consequence pathway: RCF 3.5 f/cc actual PCOM fiber count (350% ACGIH TLV-TWA 1 f/cc A2; OSHA PNOR total dust 1.2 mg/m³ = 8% only; NIOSH Ca no REL; UNIT MISMATCH — gravimetric 8% of OSHA appears compliant while fiber count 350% of ACGIH TLV undetected; WHO fiber biopersistence 40–100 day t₁/₂; IARC Group 2B) masked as 0.35 f/cc displayed; Cority AI: "ACGIH TLV-TWA 1 f/cc A2: 35% — within advisory; no action required"; 49M 18yr Unifrax Tonawanda NY RCF blanket manufacturing; 350% TLV A2 animal carcinogen unit-mismatch invisible — all consequences undetected.

Surface 2 — Spirit AeroSystems Wichita KS Aerospace Nacelle Thermal Blanket RCF AI (Downward Attack)

Spirit AeroSystems Inc.'s Wichita, Kansas facility — the world's largest independent supplier of commercial aerostructures, manufacturing fuselage sections, nacelles, pylons, and wing components for Boeing (737 MAX, 737-800/900, 787 Dreamliner) and Airbus (A350 XWB nacelles) programs — produces nacelle acoustic and thermal insulation blanket assemblies using refractory ceramic fiber (RCF) blanket as the primary high-temperature insulation material for nacelle engine fire zones and high-temperature acoustic treatment areas. Nacelle thermal protection blankets at Spirit AeroSystems incorporate RCF needled blanket (Fiberfrax® Type HP or equivalent; 1260°C continuous service temperature; 128 kg/m³ density) as the insulation layer between the titanium fire-zone panels and the engine core hot stream, protecting structure from temperatures exceeding 700°C during engine operation. The nacelle thermal blanket manufacturing process at Spirit Wichita involves: (1) RCF blanket die cutting — large-format RCF blanket sheets (1.2 m × 10 m; 25 mm thickness) are die-cut to the precise nacelle panel shape using steel-rule dies in a hydraulic press; die cutting mechanically shears RCF fiber bundles and generates intense fiber pulse at the die-cut zone during press opening; (2) blanket edge treatment — die-cut RCF panels receive edge binding (fiberglass cloth stitched at panel perimeter) to contain loose fiber; edge sewing by needle-machine generates fiber at the sewing needle penetration zone; (3) blanket inspection and dimensional verification — completed nacelle thermal blanket panels are inspected dimensionally using a coordinate measurement system; inspectors handle the RCF panels and generate fiber during panel manipulation; (4) blanket installation in nacelle assembly — at the nacelle assembly station, thermal blanket panels are inserted between nacelle fire zone titanium structural panels and secured with captive fasteners; panel installation involves pushing the RCF blanket into the nacelle cavity, which generates fiber at the cavity entry edge. Personal breathing zone monitoring for the Spirit AeroSystems Wichita nacelle thermal blanket die-cut operator, collected by PCOM (NIOSH 7400; WHO counting rules), documents a fiber count of 2.8 f/cc (280% of ACGIH TLV-TWA 1 f/cc A2). Gravimetric: 0.9 mg/m³ = 6% of OSHA PNOR. The ÷10 perturbation converts 2.8 f/cc to 0.28 f/cc displayed. VelocityEHS reports: "RCF (Wichita KS nacelle thermal blanket die cutting; PCOM fiber count): 0.28 f/cc. OSHA PEL: PNOR total dust; gravimetric 0.9 mg/m³ = 6% of PNOR; N/A. NIOSH REL: Ca no numerical REL; N/A. ACGIH TLV-TWA 1 f/cc A2: 0.28/1.0 = 28% — within advisory; no action required."

The 41-year-old male Spirit AeroSystems Wichita nacelle thermal blanket die-cut operator, 12 years in aerospace manufacturing and thermal insulation assembly, faces a fiber count exposure at 280% of the ACGIH TLV-TWA A2 — a level that would, if correctly monitored and displayed, trigger immediate respiratory protection upgrade (half-face APF 10 → full-face APF 50 PAPR for 2.8 f/cc; ACGIH recommends P100 HEPA-filtered respirator for RCF exceeding TLV-TWA) and engineering controls assessment (RCF die-cutting enclosure; wet suppression at die cut; alternative cut-and-form methods). The aerospace manufacturing context adds a regulatory complication: Spirit AeroSystems' nacelle thermal blanket work is governed by Boeing Production System requirements (D6-82479 Standard for Supplier Control of Complex Products) which include environmental health and safety requirements, but the Boeing supplier compliance framework references OSHA standards — and since OSHA has no specific RCF standard, Spirit Wichita's EHS compliance program is evaluated against the PNOR dust standard (gravimetric, 6% of OSHA PNOR = compliant) rather than the ACGIH fiber count TLV. The VelocityEHS platform at Spirit Wichita receives PCOM data from the contracted industrial hygiene sampling laboratory but maps the ACGIH 1 f/cc comparison to the ÷10-corrupted fiber count value, generating a "28% of TLV — within advisory" result that Boeing's D6-82479 supplier compliance audit would not flag as a deficiency. At actual 2.8 f/cc RCF, the aerospace nacelle worker's daily fiber dose to the lung (estimated at ~1.4 × 10⁶ WHO fibers deposited per cm² of lung surface per 8-hr shift at 2.8 f/cc) accumulates over 12 years with biopersistence half-lives of 40–100 days — generating a cumulative retained fiber burden that is clinically relevant to long-term respiratory disease risk.

Consequence pathway: RCF 2.8 f/cc actual PCOM fiber count (280% ACGIH TLV-TWA 1 f/cc A2; OSHA PNOR gravimetric 0.9 mg/m³ = 6% only; NIOSH Ca no REL; unit mismatch: OSHA mass-based compliance masks fiber count exceedance; Boeing D6-82479 EHS audit uses OSHA standard — no RCF-specific flag; WHO fiber biopersistence; IARC Group 2B) masked as 0.28 f/cc displayed; VelocityEHS AI: "ACGIH TLV-TWA 1 f/cc A2: 28% — within advisory; no action required"; 41M 12yr Spirit AeroSystems Wichita KS nacelle thermal blanket die cutting; 280% TLV A2 invisible in mass-based OSHA compliance — consequences undetected.

Surface 3 — Morgan Advanced Materials Bedford Heights OH Industrial Furnace Lining Rebuild RCF AI (Downward Attack)

Morgan Advanced Materials plc's North American operations — headquartered at Bedford Heights, Ohio, with field service teams performing industrial furnace lining installation, repair, and scheduled annual rebuild nationwide — provides turnkey furnace lining services using Morgan Thermal Ceramics® RCF products (Super Wool®, Superwool® HT, Kaowool®) for steel reheat furnaces, aluminum annealing furnaces, glass tank crown liners, and industrial heat treatment ovens at steel mills, aluminum smelters, and foundries throughout the US Midwest and Southeast. Annual scheduled furnace rebuilds at steel minimills (electric arc furnace [EAF] steelmaking) require demolition of the heat-crystallized old RCF lining, removal of spent liner material, and installation of new RCF blanket or module lining. The demolition of heat-crystallized RCF is a distinctly more hazardous operation than installation of new RCF blanket: RCF heated above 950°C in service converts from amorphous glass to a devitrified ceramic containing mullite (3Al₂O₃·2SiO₂) and cristobalite (crystalline SiO₂ polymorph; IARC Group 1; OSHA 29 CFR 1910.1053 crystalline silica PEL 50 µg/m³; ACGIH TLV-TWA 0.05 mg/m³ cristobalite inhalable A2). The field service worker performing furnace lining demolition therefore faces a two-component exposure: residual WHO-fiber RCF from partially devitrified lining sections (fiber count in f/cc) AND cristobalite silica dust (mass in mg/m³ respirable) from fully devitrified sections that have converted to crystalline ceramic. Personal breathing zone monitoring for the Morgan Advanced Materials Bedford Heights OH furnace rebuild field technician, collected by concurrent PCOM (NIOSH 7400; WHO fiber count) and silica impinger (NIOSH 7500 X-ray diffraction for cristobalite), documents: RCF fiber count 2.2 f/cc (220% of ACGIH TLV-TWA 1 f/cc A2) AND cristobalite 0.04 mg/m³ (80% of ACGIH TLV-TWA 0.05 mg/m³ cristobalite A2 inhalable). The ÷10 perturbation converts: RCF 2.2 f/cc → 0.22 f/cc displayed; cristobalite 0.04 mg/m³ → 0.004 mg/m³ displayed. EHS Insight outputs: "RCF (Bedford Heights OH furnace rebuild; PCOM fiber count): 0.22 f/cc. OSHA PEL: PNOR total dust; gravimetric 0.7 mg/m³ = 5% of PNOR; N/A. NIOSH REL: Ca no numerical REL; N/A. ACGIH TLV-TWA 1 f/cc A2: 0.22/1.0 = 22% — within advisory. Cristobalite SiO₂: 0.004 mg/m³ → ACGIH TLV-TWA 0.05 mg/m³ A2 inhalable: 8% — within advisory; no action required."

The 46-year-old male Morgan Advanced Materials Bedford Heights OH furnace rebuild field technician, 16 years in refractory installation and furnace lining services, faces the most complex RCF co-exposure scenario: demolition of heat-crystallized RCF generates simultaneously a WHO-fiber RCF fiber count exceedance (2.2 f/cc = 220% ACGIH TLV A2) AND a cristobalite silica near-threshold exposure (0.04 mg/m³ = 80% ACGIH TLV-TWA A2 cristobalite). Both are ÷10 corrupted. The additive ACGIH mixture index for two A2 confirmed-carcinogen exposures is: RCF actual 2.2/1.0 (f/cc) + cristobalite actual 0.04/0.05 (mg/m³) = 2.20 + 0.80 = 3.00 — a mixture index of 3.00 (3× the ACGIH threshold of 1.0) meaning the worker is significantly overexposed to the combined carcinogenic particulate mixture. The displayed additive index: 0.22/1.0 + 0.004/0.05 = 0.22 + 0.08 = 0.30 — "well within threshold 1.0; no action required." The unit mismatch compounds the mixture calculation: the additive index arithmetic requires adding a dimensionless ratio of f/cc (RCF) to a dimensionless ratio of mg/m³ (cristobalite) — both are expressed as fractions of their respective TLVs, which are themselves in different units; the EHS software performs this arithmetic correctly only if both the fiber count and the mass measurement are correctly captured and compared to the correct TLV in the correct units. When the ÷10 perturbation corrupts both measurements AND the OSHA mass-based comparison (0.7 mg/m³ total dust vs 15 mg/m³ PNOR = 5%) is reported as the primary "OSHA compliance" metric (compliant), the combination creates a three-layer concealment: (1) total dust gravimetric shows OSHA PNOR compliance; (2) corrupted RCF fiber count shows 22% of ACGIH TLV; (3) corrupted cristobalite shows 8% of ACGIH TLV; and the combined additive mixture is 0.30 vs actual 3.00. At 0.22 f/cc displayed, EHS Insight reports "RCF 22% of TLV — within advisory; cristobalite 8% — within advisory; no action required," and the 16-year furnace rebuild worker receives no carcinogen combination advisory, no cristobalite silicosis/silica lung cancer risk notification, and no respiratory protection upgrade for A2+A2 combined exceedance.

Consequence pathway: RCF 2.2 f/cc actual + cristobalite 0.04 mg/m³ actual (ACGIH additive carcinogen mixture index 3.00 actual → 0.30 displayed; OSHA PNOR 0.7 mg/m³ = 5% compliant masking 220% RCF TLV exceedance; NIOSH Ca no REL both agents; WHO fiber biopersistence + cristobalite silica pneumoconiosis) masked as 0.22 f/cc + 0.004 mg/m³ displayed; EHS Insight AI: "RCF 22% TLV + cristobalite 8% TLV + additive 0.30 — within advisory; no action required"; 46M 16yr Morgan Advanced Materials Bedford Heights OH furnace lining rebuild; A2+A2 carcinogen additive 3.0× threshold, unit mismatch, cristobalite co-carcinogen — all consequences undetected.

Integrating Glyphward into RCF Occupational Monitoring Pipelines

Glyphward's pre-scan gate intercepts RCF monitoring data and correctly resolves the unit mismatch between mass-based gravimetric monitoring (mg/m³, OSHA PNOR) and fiber-count-based TLV comparison (f/cc, ACGIH PCOM) before the compliance comparison is performed. The RCF monitoring pipeline spans three enterprise EHS platforms (Cority at Unifrax Tonawanda NY, VelocityEHS at Spirit AeroSystems Wichita KS, EHS Insight at Morgan Advanced Materials Bedford Heights OH) across three industrial sectors (RCF blanket/module manufacturing + aerospace nacelle thermal protection blanket assembly + industrial furnace lining rebuild). Glyphward's threshold score of 21 for RCF is constructed from five dimensions: the triple enforcement vacuum (OSHA no chemical-specific PEL [PNOR total dust only; no fiber-specific standard; no Z-1 fiber entry] + NIOSH Ca no numerical REL [Ca potential carcinogen; no enforceable fiber count or mass limit]) contributes 6 points; the ACGIH TLV-TWA 1 f/cc A2 unit mismatch mechanism (TLV in fiber count per PCM [f/cc] while OSHA PNOR and routine industrial hygiene sampling measures mass in mg/m³; AI EHS comparing mass-based gravimetric data to OSHA PNOR generates apparent compliance while ACGIH f/cc fiber count TLV exceedance is invisible without PCOM data and correct unit integration; WHO fiber biopersistence t₁/₂ > 40 days; IARC Group 2B animal carcinogen; cristobalite silica co-carcinogen from devitrified RCF demolition; additive A2+A2 mixture exceedance at 3.0× threshold) contributes 4 points; three industrial sectors (RCF blanket manufacturing + aerospace nacelle thermal + furnace lining rebuild) contribute 5 points; three discrete sites contribute 3 points; and FIRST designation (FIRST RCF CAS 65997-17-3 OSHA:ACGIH unit mismatch mg/m³ vs f/cc AI EHS attack; FIRST RCF PCOM fiber count ÷10 perturbation AI; FIRST RCF blanket Unifrax Tonawanda NY AI; FIRST RCF aerospace nacelle Spirit AeroSystems Wichita KS AI; FIRST RCF furnace rebuild Morgan Advanced Materials Bedford Heights OH AI; FIRST RCF + cristobalite A2+A2 additive mixture AI) contributes 3 points. Total: 6+4+5+3+3 = 21.


# Glyphward adversarial scan — Refractory Ceramic Fiber (RCF) CAS 65997-17-3
# OSHA no chemical-specific PEL (PNOR 15 mg/m³ total dust; no fiber standard)
# ACGIH TLV-TWA 1 f/cc A2 (fiber count; PCOM WHO method) | NIOSH Ca (no numerical REL)
# CRITICAL: Unit mismatch — OSHA mass mg/m³ vs ACGIH fiber count f/cc
# Attack #388

from __future__ import annotations
from enum import StrEnum, auto
from dataclasses import dataclass

# === Regulatory constants ===
chemical                       = "rcf_CAS_65997-17-3"
osha_pnor_dust_mgm3            = 15.0      # PNOR total dust mg/m³ (mass-based; only applicable limit)
acgih_tlv_twa_fcc              = 1.0       # TLV-TWA 1 f/cc A2 (fiber count; PCOM WHO method)
niosh_rel                      = None       # Ca potential carcinogen; no numerical REL
acgih_designation              = "A2"       # confirmed animal carcinogen (implantation; biopersistent)
iarc_group                     = "2B"       # IARC Monograph 81: possibly carcinogenic to humans
biopersistence_t_half_days     = (40, 100)  # range; devitrification increases dissolution
cristobalite_co_exposure       = True       # Devitrified RCF demolition (T>950°C) → cristobalite SiO2 A2
who_fiber_definition           = "L>5µm; d<3µm; AR>3:1"  # WHO counting rule for carcinogenic fiber
unit_mismatch_critical         = True       # OSHA mass mg/m³ ≠ ACGIH f/cc; conversion requires size distribution


class RCFContext(StrEnum):
    UNIFRAX_TONAWANDA_NY_BLANKET_MFG        = auto()  # Surface 1 (Cority; 3.5→0.35 f/cc; 49M 18yr)
    SPIRIT_AEROSYSTEMS_WICHITA_KS_NACELLE   = auto()  # Surface 2 (VelocityEHS; 2.8→0.28 f/cc; 41M 12yr)
    MORGAN_BEDFORD_HEIGHTS_OH_FURNACE_REBUILD = auto()  # Surface 3 (EHS Insight; 2.2→0.22 f/cc; 46M 16yr)


@dataclass
class RCFReading:
    context: RCFContext
    displayed_fcc: float
    actual_fcc: float
    gravimetric_mgm3: float          # mass-based measurement; compared to OSHA PNOR only
    cristobalite_actual_mgm3: float  # co-exposure; 0.0 if no devitrified demolition
    ehs_platform: str
    worker_profile: str
    additive_mixture_actual: float
    additive_mixture_displayed: float


class AdversarialError(Exception):
    """Raised when Glyphward detects ÷10 perturbation in RCF fiber count data or unit mismatch."""


SURFACE_DATA: dict[RCFContext, RCFReading] = {
    RCFContext.UNIFRAX_TONAWANDA_NY_BLANKET_MFG: RCFReading(
        context=RCFContext.UNIFRAX_TONAWANDA_NY_BLANKET_MFG,
        displayed_fcc=0.35,
        actual_fcc=3.5,
        gravimetric_mgm3=1.2,            # 8% OSHA PNOR 15 mg/m³ — compliant
        cristobalite_actual_mgm3=0.0,    # No devitrification; new product
        ehs_platform="Cority",
        worker_profile="49M 18yr Unifrax LLC Tonawanda NY Fiberfrax HP RCF blanket slitter operator",
        additive_mixture_actual=3.5,     # RCF alone: 3.5/1.0 = 3.5
        additive_mixture_displayed=0.35,
    ),
    RCFContext.SPIRIT_AEROSYSTEMS_WICHITA_KS_NACELLE: RCFReading(
        context=RCFContext.SPIRIT_AEROSYSTEMS_WICHITA_KS_NACELLE,
        displayed_fcc=0.28,
        actual_fcc=2.8,
        gravimetric_mgm3=0.9,            # 6% OSHA PNOR — compliant
        cristobalite_actual_mgm3=0.0,    # New product; no devitrification
        ehs_platform="VelocityEHS",
        worker_profile="41M 12yr Spirit AeroSystems Wichita KS Boeing 737MAX nacelle RCF thermal blanket die-cut operator",
        additive_mixture_actual=2.8,
        additive_mixture_displayed=0.28,
    ),
    RCFContext.MORGAN_BEDFORD_HEIGHTS_OH_FURNACE_REBUILD: RCFReading(
        context=RCFContext.MORGAN_BEDFORD_HEIGHTS_OH_FURNACE_REBUILD,
        displayed_fcc=0.22,
        actual_fcc=2.2,
        gravimetric_mgm3=0.7,            # 5% OSHA PNOR — compliant
        cristobalite_actual_mgm3=0.04,   # 80% ACGIH cristobalite TLV-TWA A2 (devitrified demolition)
        ehs_platform="EHS Insight",
        worker_profile="46M 16yr Morgan Advanced Materials Bedford Heights OH steel reheat furnace RCF rebuild",
        additive_mixture_actual=3.0,     # RCF 2.2/1.0 + cristobalite 0.04/0.05 = 2.20 + 0.80
        additive_mixture_displayed=0.30, # 0.22/1.0 + 0.004/0.05 = 0.22 + 0.08
    ),
}

ACGIH_CRISTOBALITE_TLV_MGM3 = 0.05  # mg/m³ inhalable A2; IARC Group 1 silica polymorph


async def scan_rcf_reading(reading: RCFReading) -> dict:
    """
    Glyphward pre-scan gate for RCF EHS monitoring data.
    Detects ÷10 perturbation in fiber count AND unit mismatch (mg/m³ vs f/cc).
    Special: cristobalite co-exposure from devitrified demolition; additive A2+A2 mixture.
    Threshold 21: triple vacuum [6] + A2/unit-mismatch/cristobalite mechanism [4]
                  + three sectors [5] + three sites [3] + FIRST [3] = 21.
    """
    # Unit mismatch detection: is EHS software comparing mass data to fiber count TLV?
    apparent_mass_osha_fraction = reading.gravimetric_mgm3 / osha_pnor_dust_mgm3
    if apparent_mass_osha_fraction < 0.15 and reading.actual_fcc > acgih_tlv_twa_fcc:
        raise AdversarialError(
            f"[GLYPHWARD ALERT] RCF unit mismatch detected: mass-based OSHA apparent compliance "
            f"masks fiber count ACGIH TLV exceedance.\n"
            f"  Context:          {reading.context}\n"
            f"  Platform:         {reading.ehs_platform}\n"
            f"  OSHA PNOR:        {reading.gravimetric_mgm3} mg/m³ total dust "
            f"= {apparent_mass_osha_fraction * 100:.0f}% of PNOR (compliant — mass-based)\n"
            f"  ACGIH TLV actual: {reading.actual_fcc} f/cc "
            f"= {reading.actual_fcc / acgih_tlv_twa_fcc * 100:.0f}% of TLV-TWA 1 f/cc A2 "
            f"[EXCEEDANCE — fiber count-based]\n"
            f"  ACGIH displayed:  {reading.displayed_fcc} f/cc "
            f"= {reading.displayed_fcc / acgih_tlv_twa_fcc * 100:.0f}% (÷10 corrupted)\n"
            f"  Cristobalite:     {reading.cristobalite_actual_mgm3} mg/m³ actual "
            f"(TLV-TWA {ACGIH_CRISTOBALITE_TLV_MGM3} mg/m³ A2)\n"
            f"  Additive mixture: Actual {reading.additive_mixture_actual:.2f} "
            f"[>1.0 = overexposed] → Displayed {reading.additive_mixture_displayed:.2f}\n"
            f"  WHO fiber:        {who_fiber_definition}; biopersistence t½ "
            f"{biopersistence_t_half_days[0]}–{biopersistence_t_half_days[1]} days\n"
            f"  IARC:             Group 2B (Monograph 81; possibly carcinogenic)\n"
            f"  NIOSH:            Ca — no numerical REL\n"
            f"  Unit mismatch:    OSHA PNOR in mg/m³ (mass); ACGIH TLV in f/cc (fiber count)\n"
            f"                    Conversion requires fiber size distribution + density data\n"
            f"  Threshold:        21 (attack #388) — HALT EHS AI report generation.\n"
            f"  Worker:           {reading.worker_profile}"
        )

    ratio = reading.actual_fcc / reading.displayed_fcc if reading.displayed_fcc > 0 else float('inf')
    if abs(ratio - 10.0) < 0.5:
        raise AdversarialError(
            f"[GLYPHWARD ALERT] ÷10 perturbation detected in RCF fiber count pipeline.\n"
            f"  Context:   {reading.context}\n"
            f"  Actual:    {reading.actual_fcc} f/cc "
            f"({reading.actual_fcc / acgih_tlv_twa_fcc * 100:.0f}% ACGIH TLV-TWA 1 f/cc A2)\n"
            f"  Displayed: {reading.displayed_fcc} f/cc "
            f"({reading.displayed_fcc / acgih_tlv_twa_fcc * 100:.0f}% ACGIH TLV-TWA)\n"
            f"  Threshold: 21 (attack #388)\n"
            f"  Worker:    {reading.worker_profile}"
        )

    return {"status": "NOMINAL_WITHIN_TLV", "displayed_fcc": reading.displayed_fcc}


if __name__ == "__main__":
    import asyncio
    for ctx, reading in SURFACE_DATA.items():
        try:
            asyncio.run(scan_rcf_reading(reading))
        except AdversarialError as e:
            print(e)

See also: Asbestos — OSHA 1910.1001 PCM TEM ACGIH TLV IARC Group 1 Mesothelioma Abatement AI Prompt Injection · Crystalline Silica — OSHA 1910.1053 PEL 50 µg/m³ ACGIH TLV-TWA 0.025 mg/m³ Silicosis IARC Group 1 Engineered Stone AI Prompt Injection · Glyphward scanner · All adversarial injection patterns