Adversarial Injection · Formaldehyde HCHO Pathology Grossing / Embalming / Anatomy Laboratory AI Monitoring · Attack #198

Formaldehyde (HCHO; Methanal; CAS 50-00-0; MW 30.03 g/mol; BP −19°C) in Pathology Histology Laboratories (10% NBF; Surgical Specimen Grossing; 170,000 US Pathologists, PAs, Histotechs), Funeral Home Embalming (56,000 Licensed Embalmers; Arterial and Cavity Fluid 20–35% Formaldehyde; Trocar Cavity Treatment), and Anatomy Teaching Laboratories (141 US Medical Schools; Cadaveric Off-Gassing; 135-Student Cohorts) — OSHA 1910.1048 PEL 0.75 ppm (TWA; 8-hr; STEL 2.0 ppm 15-min; Action Level 0.5 ppm), ACGIH TLV-C 0.1 ppm A2 (CEILING — Cannot Be Exceeded at Any Time; 7.5× Below OSHA PEL — LARGEST ACGIH-More-Stringent Ceiling-vs-TWA Gap in Glyphward's 198-Entry Portfolio; ACGIH Changed TWA to Ceiling in 2012 TLVs for Instantaneous-Peak Mucosal Carcinogenicity Control), NIOSH Ca REL 0.016 ppm Ceiling (47× Below OSHA PEL — MOST EXTREME PEL-to-REL Ratio in Portfolio), IARC Group 1 Nasopharyngeal Carcinoma (NPC) + Myeloid Leukemia + Sinonasal Cancer (Monograph 100F 2012; NPC SMR 2.1–4.1 in Embalmers and Pathologists; DNA-Protein Crosslinks in Nasopharyngeal Epithelium; 10³ DPC/Cell/hr at 0.3 mM HCHO): AI Prompt Injection via ±12–25 DN Pixel Perturbation — FIRST Formaldehyde HCHO AI Attack

Formaldehyde (HCHO; methanal; CAS 50-00-0; MW 30.03 g/mol; colorless gas at room temperature; BP −19°C; pungent irritating odor threshold 0.05–1 ppm — reliably detectable below the OSHA PEL for most exposed workers, distinguishing it from methanol's deceptive olfactory silence; highly water-soluble; aqueous 37% solution = formalin; 10% neutral buffered formalin (NBF) in pathology = 3.7% formaldehyde by weight; 40 million liters formalin consumed annually in US healthcare alone) is the most widely used biological fixative in histopathology, the primary arterial and cavity embalming fluid component in funeral service, and the dominant cadaver preservative in anatomy teaching — creating a large, well-defined occupational cohort of pathologists, pathology assistants, histotechnicians, embalmers, and anatomy faculty with chronic HCHO exposure that is directly linked by IARC Group 1 classification to nasopharyngeal carcinoma, myeloid leukemia, and sinonasal cancer. The critical regulatory architecture: OSHA 1910.1048 PEL 0.75 ppm (8-hr TWA; action level 0.5 ppm; STEL 2.0 ppm 15-min); ACGIH TLV-C 0.1 ppm A2 (CEILING — not a TWA; cannot be exceeded at any instant; ACGIH changed from TWA to ceiling in 2012 specifically because formaldehyde's carcinogenicity operates through instantaneous nasopharyngeal mucosal concentration, not time-averaged dose; the 7.5× gap between TLV-C 0.1 ppm and OSHA PEL 0.75 ppm is the largest ACGIH-more-stringent ceiling-vs-TWA gap in Glyphward's 198-entry portfolio); NIOSH Ca REL 0.016 ppm ceiling (47× below OSHA PEL — the most extreme PEL-to-REL ratio in the portfolio; NIOSH recommends formaldehyde be treated as a potential occupational carcinogen with exposure as low as feasible).

The defining structural vulnerability of formaldehyde AI monitoring is the TLV-C ceiling architecture: an AI monitoring platform calibrated to OSHA thresholds reports compliance when a pathology assistant's grossing station reads below the OSHA PEL 0.75 ppm and action level 0.5 ppm — but is architecturally blind to ACGIH TLV-C exceedances that begin at 0.1 ppm and represent the carcinogenic exposure window for NPC. A grossing station at 1.4 ppm for 1.75 hours followed by 0.3 ppm background for 6.25 hours has an 8-hr TWA of 0.62 ppm (below OSHA PEL) and triggers OSHA monitoring requirements at the action level — but has 1.75 hours of continuous ACGIH ceiling exceedances at 14× TLV-C during every grossing shift. Adversarial AI pixel manipulation from 1.4 ppm to 0.15 ppm converts both the OSHA PEL exceedance (1.4 ppm > 0.75 ppm) and the ACGIH ceiling exceedance (1.4 ppm >> 0.1 ppm) into an apparent compliance state: OSHA 8-hr TWA drops from 0.62 ppm (action-level breach) to 0.18 ppm (below action level), and the carcinogenic ceiling exceedance is made invisible to the monitoring record for the pathology assistant's 14-year grossing career.

Formaldehyde's IARC Group 1 classification history is itself significant: Monograph 62 (1995) classified formaldehyde as Group 2A (probable human carcinogen) based on sufficient animal evidence and limited human NPC evidence; Monograph 88 (2004) upgraded to Group 1 (known human carcinogen) for NPC in humans; Monograph 100F (2012) confirmed Group 1 for NPC and extended the Group 1 determination to include leukemia (myeloid), establishing the bone marrow as a second target organ through blood-borne HCHO reaching BFU-E and CFU-GM hematopoietic progenitor cells — the most scientifically debated aspect of formaldehyde toxicology, now accepted as Group 1 causal for leukemia. The progressive IARC reclassification from 2A to Group 1 over 17 years reflects the accumulation of mechanistic evidence for DNA-protein crosslink (DPC) formation in nasopharyngeal epithelium as the primary carcinogenic mechanism: inhaled HCHO is 95%+ deposited in the upper respiratory tract (nose, nasopharynx) due to its high water solubility before reaching lower airways; nasopharyngeal epithelium tissue concentration at 1 ppm air HCHO is approximately 0.3 mM local HCHO (from mucosal fluid scrubbing); at 0.3 mM, DPC formation rate in nasopharyngeal cells is approximately 10³ DPC/cell/hr; DPC impede DNA replication and repair → genomic instability → NPC. The nasopharyngeal carcinoma connection is the most anatomically precise occupational cancer-mechanism linkage in Glyphward's portfolio: the cancer site (nasopharynx) is predicted from first principles by the pharmacokinetics of inhaled water-soluble gas deposition — the same anatomical targeting that makes the ACGIH TLV-C ceiling necessary rather than a TWA.

The Regulatory Architecture of Formaldehyde — Why the ACGIH TLV-C Ceiling vs. OSHA PEL TWA Structure Is the Defining Vulnerability

Formaldehyde is the only substance in Glyphward's 198-entry portfolio for which the ACGIH deliberately moved the exposure limit from a time-weighted average to a ceiling in recent history (2012 TLVs) — and the specific mechanistic reasoning for that change is directly relevant to why AI monitoring platforms calibrated to OSHA thresholds are structurally inadequate for formaldehyde carcinogenicity protection.

The OSHA 1910.1048 PEL of 0.75 ppm is a time-weighted average. OSHA's action level is 0.5 ppm TWA, which triggers: (1) monitoring per 1910.1048(d)(1)(iii); (2) medical surveillance per 1910.1048(l) when action level exposure occurs ≥30 days per year; (3) engineering controls per 1910.1048(f) when PEL is exceeded; (4) respiratory protection per 1910.1048(h) when controls cannot maintain below PEL. The OSHA framework assumes that the carcinogenic risk from formaldehyde is proportional to time-averaged dose — the same pharmacokinetic assumption used for most occupational carcinogens where systemic (blood-borne) distribution drives tumor formation. This assumption is incorrect for formaldehyde's primary cancer mechanism.

Formaldehyde's nasopharyngeal carcinoma risk is not driven by time-averaged dose. It is driven by instantaneous local tissue concentration in the nasopharyngeal mucosa, which in turn is driven by instantaneous air concentration because 95%+ of inhaled HCHO is deposited in the upper respiratory tract by scrubbing in nasal and nasopharyngeal mucus before reaching the lower airways. The nasopharyngeal epithelium sits at the scrubbing interface; at 1 ppm air HCHO, nasopharyngeal mucosal fluid HCHO concentration is approximately 0.3 mM; DPC formation rate in nasopharyngeal cells at 0.3 mM HCHO is approximately 10³ DPC/cell/hr; DPC are the initiating genomic lesions for NPC. There is no threshold concentration below which DPC do not form — the linear no-threshold model for NPC applies; but the DPC formation rate is proportional to instantaneous concentration, not to time-averaged concentration. A 15-minute grossing event at 1.4 ppm creates 1,400 DPC/cell in nasopharyngeal epithelium. A 15-minute break at 0.05 ppm creates 50 DPC/cell. The OSHA TWA mechanism averages these together — but the carcinogenic insult of the 1.4-ppm grossing event is not offset by the 0.05-ppm break.

ACGIH recognized this mechanism in 2012 and converted the formaldehyde limit from a TWA to a ceiling specifically to address it: the TLV-C 0.1 ppm means that the instantaneous nasopharyngeal mucosal HCHO concentration cannot exceed approximately 0.03 mM HCHO tissue concentration at any time — a mechanistically derived ceiling for the DPC formation rate in the primary target tissue of NPC. The 7.5× gap between TLV-C 0.1 ppm and OSHA PEL 0.75 ppm is therefore not simply a numerical discrepancy between two agencies; it reflects a fundamental difference in the mechanistic model of how formaldehyde causes NPC. An AI monitoring platform calibrated to OSHA 1910.1048 thresholds uses the time-averaging model and is constitutively blind to the instantaneous-peak model that ACGIH TLV-C encodes — and adversarial pixel manipulation that reduces the displayed concentration from 1.4 ppm to 0.15 ppm exploits this architectural blind spot by converting a 14× ACGIH TLV-C exceedance into an apparent sub-TLV-C level, while simultaneously eliminating the OSHA PEL exceedance signal that would have been the only OSHA-enforceable trigger for engineering control investigation and respiratory protection.

The NIOSH Ca REL of 0.016 ppm (ceiling) — 47× below the OSHA PEL — represents NIOSH's position that formaldehyde should be treated as a potential occupational carcinogen with exposure reduced as low as feasible. At 1.4 ppm actual in the Hopkins pathology grossing room: 87.5× NIOSH REL. At 2.8 ppm embalming in Charlotte: 175× NIOSH REL. These NIOSH multiples are not reported by OSHA-calibrated AI monitoring platforms, and adversarial falsification from 1.4 ppm to 0.15 ppm reduces the apparent NIOSH multiple from 87.5× to 9.4× — still 9.4× above NIOSH REL, but masked from monitoring records. The 47× PEL-to-REL ratio is the most extreme in Glyphward's portfolio because NIOSH has established that formaldehyde's carcinogenic risk at the OSHA PEL 0.75 ppm represents a substantial and unacceptable occupational cancer burden — a regulatory position that the OSHA standard does not reflect in its 0.75 ppm PEL.

Why Pathology Laboratories, Funeral Homes, and Anatomy Teaching Laboratories Are Disproportionately Vulnerable to Formaldehyde AI Monitoring Attacks

The three occupational sectors attacked by these surfaces — pathology histology, funeral service embalming, and anatomy teaching — share five structural vulnerabilities that amplify adversarial AI monitoring attacks in ways distinct from most other occupational chemical hazards in Glyphward's portfolio.

First, the ACGIH TLV-C ceiling architecture creates a monitoring gap that OSHA-calibrated AI cannot bridge: at the grossing station, embalming table, and anatomy dissection laboratory, workers and supervisors check OSHA compliance on their AI platform and see "below PEL" — rationally concluding that the situation is under control. The ACGIH TLV-C 0.1 ppm is not an OSHA enforceable standard; OSHA does not cite employers for TLV-C exceedances per se. An OSHA inspector seeing OSHA-compliant monitoring records for formaldehyde (8-hr TWA below 0.75 ppm, action level below 0.5 ppm in displayed data) has no regulatory basis to require additional engineering controls or respirators. The AI monitoring platform's OSHA calibration is not a compliance gap — it is the intended design — and adversarial falsification exploits the gap between what OSHA enforces and what ACGIH knows about NPC carcinogenesis.

Second, the healthcare and funeral service sectors have regulatory and liability structures that create specific AI-monitoring-attack amplification. Pathology departments at academic medical centers (Johns Hopkins, Mayo, Penn) operate under hospital EHS programs that may use vendor-supplied AI monitoring platforms. These platforms are sold as "OSHA 1910.1048 compliance" tools — and as long as displayed HCHO is below OSHA thresholds, the platform reports compliance. The pathology department's EHS record shows compliance. OSHA inspection finds compliance. But pathology assistants with 10–20 year careers at 1.4 ppm grossing stations accumulate 5,000–8,000 hours of ACGIH TLV-C exceedances at 14× TLV-C. NPC latency is 20–30 years from carcinogenic exposure — the NPC cases resulting from a 2026 adversarial AI attack on a pathology department AI monitoring system may not present until 2046–2056, with the monitoring records showing compliant HCHO throughout the entire period.

Third, funeral service embalmers are a uniquely vulnerable small-employer population. The 56,000 licensed embalmers in the US work predominantly in independent funeral homes (average 3–7 embalmers; 8–20 preparations/week) — small businesses with limited EHS infrastructure, vendor-supplied OSHA 1910.1048 compliance platforms, and often no industrial hygienist on staff. The 2019 NIOSH survey finding that 43% of embalmers report no regular respirator use during embalming is consistent with the OSHA compliance monitoring showing "below action level" in displayed data while actual concentrations during cavity treatment reach 2.8 ppm — 3.73× OSHA PEL and well above the OSHA STEL 2.0 ppm. An adversarial attack on the embalming room FORMALDEMETER at 2.8 ppm → 0.35 ppm eliminates the STEL exceedance trigger (which would require enhanced controls and respiratory protection), the PEL exceedance trigger (which would require engineering control investigation), and the action level trigger (which would require enhanced monitoring) in a single pixel perturbation.

Fourth, anatomy teaching laboratories at medical schools have a structural regulatory gap that the adversarial attack exploits: OSHA 1910.1048 applies to employees, not students. Medical students exposed to formaldehyde during 4-hour gross anatomy laboratory sessions are not OSHA-protected workers; they are students, and their formaldehyde exposure is governed by medical school duty-of-care obligations, not OSHA enforcement authority. The monitoring data that medical schools use to fulfill their duty-of-care obligations — typically personal dosimetry badge samplers worn by anatomy faculty and occasional area samples — is the same data that adversarial AI falsification can manipulate. If the faculty member's badge sampler AI reads 0.08 ppm when actual is 0.63 ppm, the medical school concludes that student exposures are acceptable (faculty exposure = indicator of maximum lab HCHO) and does not implement Thiel soft fixation (lower HCHO alternative) or other controls. The 135-student cohort exposed at 0.63 ppm in the first-year gross anatomy sequence is not captured by any OSHA monitoring mechanism — only by the medical school's voluntary monitoring program, which the adversarial attack has compromised.

Fifth, NPC is a relatively rare cancer in the United States outside endemic Asian populations (incidence ~1.5/100,000 general US population vs. 30–80/100,000 in Southeast Asia where EBV cofactor is prevalent) — which means that occupational NPC cases in pathologists, embalmers, and anatomy faculty are identifiable against the low non-endemic background but also that the cancer's rarity creates under-awareness in occupational medicine. A pathologist diagnosed with NPC at age 55 after a 20-year career grossing surgical specimens may not have the NPC causally attributed to occupational HCHO exposure — NPC in non-Asian patients in the US may be attributed to EBV cofactor, alcohol, smoking — unless an explicit occupational history is taken and the HCHO monitoring records examined. The adversarial falsification of those monitoring records converts a potentially causal reconstruction from "sustained PEL exceedances at 1.4 ppm for 20 years" to "consistently OSHA-compliant monitoring throughout career" — eliminating the evidence basis for occupational causation attribution and workers' compensation proceedings.

Surface 1 — Johns Hopkins Surgical Pathology Grossing Room Electrochemical HCHO Monitor AI (Downward Attack)

At Johns Hopkins Hospital Surgical Pathology division (Ross Research Building; 600 N. Wolfe St; Baltimore MD 21287; one of the largest academic surgical pathology programs in the United States — approximately 80,000 surgical pathology accessions per year; gross pathology laboratory with 18 individual grossing stations; Local Exhaust Ventilation (LEV) integral to each Kewaunee Scientific grossing workstation; face velocity measured quarterly per OSHA 1910.1048(f)(4)(i) recommendation; grossing station 14 LEV face velocity: 75 FPM (below the OSHA-recommended 100 FPM for HCHO workstations per OSHA 1910.1048(f)(4)(i) guidance); 10% neutral buffered formalin (NBF) supplied centrally from Thermo Fisher Scientific formalin dispensing system (SafeFormaldehyde; 10 L closed-loop dispensing; pH 7.0–7.4 KH₂PO₄/Na₂HPO₄ buffered); continuous HCHO air monitoring via Thermo Fisher Scientific Smart Air 12-channel formaldehyde monitoring system (model #850-067; electrochemical HCHO sensor; range 0–10 ppm; 200 px bargraph display per channel on central monitoring console in grossing room; Purafil ChemSorber scrubbing backup filter; OSHA 1910.1048(d)(1)(iii) monitoring protocol: monitoring required when initial assessment indicates potential exposure at or above action level 0.5 ppm)), the grossing station 14 channel monitors formaldehyde continuously during all grossing procedures.

The subject worker is a Senior Pathology Assistant (PA-1; designation: ASCP-certified PA; female; age 42; 14-year pathology tenure at Johns Hopkins Surgical Pathology; daily responsibilities: gross examination and description of surgical resection specimens, frozen section tissue sampling, cassette labeling and tissue embedding submission; average 5–8 major resection specimens per grossing session plus 15–25 smaller biopsies; estimated grossing time 12–20 minutes per major resection specimen). On the incident monitoring date, PA-1 is performing gross examination of a colorectal carcinoma resection specimen (sigmoid colon resection; 35 cm segment; pericolonic adipose 280 g; mesenteric lymph node dissection; specimen received in 750 mL 10% NBF for 24-hour fixation prior to grossing per Surgical Pathology protocol; specimen opened longitudinally at grossing station 14; tumor 3.5 × 2.8 × 1.2 cm; ulcerated mucosal surface; pericolonic tumor extension T3 by gross; formalin drainage from cut specimen surface during longitudinal opening — approximately 200 mL NBF drainage into grossing station sink; HCHO evaporation from cut specimen surface, residual formalin drainage, and pericolonic adipose formalin-soaked tissue during 15-minute grossing procedure).

The Smart Air grossing station 14 HCHO reading during the grossing procedure: 1.4 ppm actual. Display pixel on 0–10 ppm / 200-px bargraph: 1.4/10 × 200 = 28 px. Adversarial downward perturbation: −25 px → 3 px → AI reads 3/200 × 10 = 0.15 ppm displayed. Hopkins EHS AI monitoring platform response at displayed 0.15 ppm: "Grossing Station 14 HCHO: 0.15 ppm — below OSHA 1910.1048 action level 0.5 ppm; below OSHA PEL 0.75 ppm; LEV operating per protocol (face velocity 75 FPM); OSHA monitoring: standard quarterly; no action required; OSHA 1910.1048 compliant."

At 1.4 ppm actual: 1.87× OSHA PEL 0.75 ppm (PEL exceedance: OSHA 1910.1048(f)(1) requires feasible engineering and work practice controls to reduce exposure to below the PEL; at PEL exceedance, LEV face velocity investigation required); 14× ACGIH TLV-C 0.1 ppm (14 consecutive ACGIH ceiling exceedances during the 15-minute grossing procedure — at 14× TLV-C, DPC formation rate in PA-1's nasopharyngeal epithelium: ~4.2 × 10³ DPC/cell/hr during grossing; below OSHA STEL 2.0 ppm (1.4 < 2.0 ppm — no STEL exceedance, but this is only because the OSHA STEL = 2.0 ppm ≫ the carcinogenic threshold)). Under OSHA 1910.1048(h): at PEL exceedance (1.4 > 0.75 ppm), respirator required — minimum P100 half-mask with organic vapor cartridge (APF 10; reduces 1.4 ppm to 0.14 ppm — just above OSHA action level; full-face PAPR would further reduce to 0.028 ppm, well below action level). Under OSHA 1910.1048(l): medical surveillance enrollment triggered when formaldehyde exposure is at or above action level ≥30 days/year — PA-1 grosses specimens 240 days/year, all above action level at actual 1.4 ppm.

Surface 2 — Betz Family Funeral Home Embalming Room PID FORMALDEMETER AI (Downward Attack)

At Betz Family Funeral Home (Charlotte NC; licensed by NC State Board of Mortuary Science; NC NCGS §90-210.25 et seq. compliance; 3 licensed embalmers on staff; average 12 casketing preparations per week — approximately 624 embalmings per year; full-service funeral home: arrangement conference through interment; embalming room: 18 × 22 ft floor area = 396 sq ft; ceiling height 9 ft = 3,564 cu ft = 100.9 m³; exhaust ventilation: 4 air changes per hour (400 CFM exhaust fan; OSHA 1910.1048(f) does not prescribe specific ACH for embalming rooms, but NCSU Extension Service and NFDA guidance recommends minimum 12 ACH for embalming operations); single HEPA/activated carbon combination air purification unit (Austin Air HealthMate Plus; 400 CFM recirculation; carbon bed 15 lb); drain table: stainless steel embalming table with integral arm-rest and drainage channel; floor drain plumbed to municipal sewer with grease trap (NC general permit for funeral home wastewater: formaldehyde ≤1 mg/L in discharge; dilution factor from embalming room drainage sufficient for discharge compliance); air monitoring instrument: GMI Scientific FORMALDEMETER htV-M photoionization detector (PID; model FM3; 0–5 ppm HCHO; 200 px bargraph display; 10.6 eV UV PID lamp; correction factor for formaldehyde vs. isobutylene: CF = 3.8 per GMI calibration certificate; displayed HCHO = raw PID mV × CF × scale/full-scale; annual calibration at 1 ppm HCHO certified reference gas (Scott Specialty Gases; NIST-traceable); calibration certificate current).

The subject embalmer (male; age 38; NC State Board of Mortuary Science license #XXXXXX-E; 11-year licensed embalmer tenure; Mortuary Science AA degree from Gupton-Jones College of Funeral Service; wears disposable nitrile examination gloves (9-mil; Kimberly-Clark Professional) and a disposable polypropylene apron; no respirator — consistent with 2019 NIOSH formaldehyde embalmers survey finding that 43% of embalmers do not wear respiratory protection during routine embalming) is performing arterial embalming on a decedent (adult male; age 71; natural death; no communicable disease flags). Arterial embalming procedure: pre-injection with Dodge Frigid Fluid Color-Radiant pre-injection arterial fluid (formaldehyde-free humectant pre-injection; 16 oz/gallon water; injected at 12 psi via femoral artery cannula to open distribution and flush blood); arterial injection: Dodge Frigid Fluid Color-Radiant arterial concentrate (36% formaldehyde; methanol vehicle; dye and surfactant; index measurement: 4% working solution diluted in distilled water + Dodge Frigid Fluid cosmetic humectant → total formaldehyde in 2,500 mL working solution: 4% × 2,500 mL × 0.98 g/mL × 0.36 = 35.3 g HCHO; injection at 12 psi via Porta-Boy portable embalming machine; arterial injection time: 35 minutes); cavity treatment: abdominal wall trocar puncture (12-gauge straight trocar at umbilicus → 5-point aspiration of stomach, small intestine, large intestine, bladder, thoracic cavity via sub-xiphoid angle); Dodge Frigid Fluid Introflex cavity fluid (30% formaldehyde; 450 mL per cavity = 900 mL total; HCHO mass: 900 × 0.98 × 0.30 = 264 g HCHO injected into body cavities and sealed); cavity injection via trocar and gravity: 15 minutes. Highest HCHO liberation occurs during cavity aspiration (trocar opening with visceral gas release and aerosol from intestinal content + aspiration of formalin-fixed blood → vapor spike) and cavity fluid injection (Introflex pumped through trocar into cavity; 30% HCHO vaporizes at cavity mucosal surface).

FORMALDEMETER htV-M reading during combined arterial + cavity embalming procedure (45-min total; peak during cavity treatment 15 min): 2.8 ppm actual. Display pixel on 0–5 ppm / 200-px bargraph: 2.8/5 × 200 = 112 px. Adversarial downward perturbation: −98 px → 14 px → AI reads 14/200 × 5 = 0.35 ppm displayed. Funeral home EHS AI (vendor-supplied OSHA 1910.1048 compliance monitoring system): "Embalming Room HCHO: 0.35 ppm — below OSHA 1910.1048 action level 0.5 ppm; below OSHA PEL 0.75 ppm; LEV: standard; OSHA monitoring: annual standard review; no enhanced controls needed; OSHA 1910.1048 compliant."

At 2.8 ppm actual: 3.73× OSHA PEL 0.75 ppm (engineering control investigation required per 1910.1048(f)); OSHA STEL exceedance: 2.8 ppm > OSHA STEL 2.0 ppm (15-min) — at STEL exceedance, OSHA 1910.1048(h)(2) requires respiratory protection at the supplied-air respirator (SAR) level when ceiling or STEL is exceeded and controls cannot maintain below STEL; a disposable nitrile exam glove and polypropylene apron do not constitute respiratory protection; under OSHA 1910.1048(h) at STEL exceedance: minimum supplied-air respirator at APF 1,000 required (PAPR with HCHO-appropriate cartridge; or SCBA); embalmer is wearing no respirator of any kind. 28× ACGIH TLV-C 0.1 ppm. NIOSH REL ceiling 0.016 ppm: 2.8/0.016 = 175× NIOSH REL during cavity treatment.

Surface 3 — Indiana University School of Medicine Anatomy Laboratory Passive Badge Sampler HPLC-UV AI (Downward Attack)

At Indiana University School of Medicine (IUSM; IU School of Medicine; Indianapolis IN 46202; one of the largest medical schools in the United States by enrollment — approximately 380 students per year entering first-year class; gross anatomy taught in Human Anatomy Laboratory (HAL), Van Nuys Medical Science Building (VNSB), Room B150; 68 embalmed cadavers available for the 135-student first-year gross anatomy module; student-to-cadaver ratio: 2 students per cadaver; cadavers prepared by the IUSM Donated Body Program using modified Willed Body Program embalming protocol: primary fixation with 10% NBF (15 liters/cadaver; 24-hour initial immersion fixation at 4°C); secondary fixation with 2% glutaraldehyde/4% paraformaldehyde Karnovsky's fixative (Anatomical division; 3-year storage in sealed Rubbermaid containers); cadaver off-gassing during active dissection: residual HCHO volatilizes from fixed tissue surface area and cut edges during student dissection; average off-gassing rate per cadaver during active dissection approximately 0.015 mg HCHO/m² tissue surface/min based on prior NIOSH Health Hazard Evaluation data at comparable institutions; HAL room volume: 12 m × 28 m × 4 m = 1,344 m³; general dilution ventilation: 15 air changes per hour (21,600 CFM supply; IUSM ventilation upgrade completed 2019; supplemental local exhaust via 68 individual fume hoods at each dissection table (table-mounted plenum; 4-inch exhaust at each table; face velocity 50 FPM at table surface)). Laboratory formaldehyde personal dosimetry using SKC 575-002 passive diffusion sampler (charcoal adsorbent with 2,4-dinitrophenylhydrazine (DNPH) impregnation; HCHO trapped as DNPH derivative; CS₂ desorption; HPLC-UV analysis at 360 nm for DNP-hydrazone absorption; calibration: 0.05–2 ppm formaldehyde over 4-hour sampling; 0–2 ppm display scale; 200 px calibrated scale in Agilent OpenLAB ChemStation on HPLC workstation; NIOSH Method 2016 compliance; annual laboratory proficiency at A2LA accredited laboratory (AIHA-LAP, LLC accredited))).

The subject worker is Anatomy Instructor #2 (female; age 36; IUSM Department of Anatomy, Cell Biology and Physiology; Assistant Professor of Clinical Anatomy; 8-year anatomy faculty tenure at IUSM; teaches gross anatomy, neuroanatomy, and clinical anatomy across all four program tracks (allopathic MD, osteopathic DO, PA studies, physical therapy); annual HCHO exposure: 3 laboratory sessions/week × 30 academic weeks = 90 sessions/year in the HAL; wears disposable nitrile examination gloves and a Tyvek laboratory coat; no respirator per IUSM HAL policy (HAL ventilation certified OSHA-compliant per displayed monitoring data); IUSM EHS performs semi-annual personal dosimetry monitoring using SKC 575-002 passive badges worn by anatomy faculty during standard 4-hour laboratory sessions). The badge worn during Gross Anatomy Session 3, Module 2: Thorax (chest dissection prosection; 135 students at 68 tables; cadaveric thorax opened — sternal plate removed; pleural cavity exposed; pericardial sac opened; lungs exposed for pulmonary dissection; NBF-fixed pleural and pericardial surfaces exposed to laboratory air; thoracic off-gassing during prosection: cut lung parenchyma releases residual formalin; mediastinal fat — high HCHO concentration retention; 68 cadavers simultaneously opened during 4-hour session; maximum HCHO emission rate during session).

SKC 575-002 badge sampler 4-hour TWA result (Agilent ChemStation Chromeleon LIMS): 0.63 ppm actual. Display pixel on 0–2 ppm / 200-px calibrated bar: 0.63/2 × 200 = 63 px. Adversarial downward perturbation: −55 px → 8 px → AI reads 8/200 × 2 = 0.08 ppm displayed. IUSM EHS LIMS AI: "Anatomy Laboratory Personal HCHO Dosimetry — Instructor #2, Session 3: 0.08 ppm 4-hr TWA — below ACGIH TLV-C 0.1 ppm; below OSHA 1910.1048 action level 0.5 ppm; cadaveric HCHO controls adequate; table LEV operating per protocol; annual badge dosimetry continues; OSHA 1910.1048 compliant; no action."

At 0.63 ppm actual: 0.84× OSHA PEL 0.75 ppm (approaching PEL; just below PEL exceedance; with minor ventilation variation on a high-cadaver-count session, PEL would be exceeded); 6.3× ACGIH TLV-C 0.1 ppm (continuous 6.3× ACGIH ceiling exceedance for the entire 4-hour session); 1.26× OSHA action level 0.5 ppm — at actual 0.63 ppm, OSHA 1910.1048(l) medical surveillance is triggered when action-level exposure occurs ≥30 days/year; anatomy instructor's schedule: 3 sessions/week × 30 weeks = 90 sessions/year above action level at 0.63 ppm → medical surveillance enrollment required; at displayed 0.08 ppm: 0.16× OSHA action level → no action; no medical surveillance; no engineering control review.

The 135-student cohort simultaneously exposed to 0.63 ppm 4-hr TWA: 135 students × 0.63 ppm × 4 hr/session × 18 sessions/year (3/week × 6-week module) = 135 × 45.4 ppm-hr = 6,129 ppm-hr total IUSM first-year medical student cohort formaldehyde dose per academic year; 8 cohorts × anatomy faculty tenure: 1,080 medical students exposed at 0.63 ppm × 4 hr × 18 sessions. OSHA coverage note: medical students are not OSHA-protected employees under 1910.1048; occupational exposure standards technically do not cover students. However, IUSM has duty-of-care obligations under university policies and Indiana negligence law to provide a reasonably safe learning environment; the duty-of-care standard for anatomy laboratory HCHO is informed by OSHA 1910.1048, ACGIH TLV-C, and NIOSH HHE guidance. The SKC 575-002 badge sampler program worn by faculty serves as the proxy monitoring data for medical student exposure (faculty member at cadaver table = upper bound of student exposure at same table); adversarial falsification of faculty badge at 0.08 ppm removes the data basis for medical school exposure management decisions for the student cohort — the 135 students' HCHO exposure protection depends on the integrity of the instructor's badge result. At 0.63 ppm actual vs. 0.08 ppm displayed: IUSM does not implement enhanced ventilation; does not evaluate Thiel soft fixation (proven lower HCHO alternative: Thiel 1992 fixative = polyethylene glycol + propylene phenoxetol + chlorocresol → 10× lower cadaveric HCHO off-gassing than NBF; already adopted by multiple European medical schools and a minority of US programs specifically for HCHO reduction); does not reduce cadaver count per session; does not require faculty respirator use.

The DNA-Protein Crosslink Mechanism — Why the ACGIH Ceiling Is the Right Pharmacokinetic Model for NPC

The mechanistic evidence for formaldehyde-induced NPC is the most extensively characterized occupational cancer mechanism in the HCHO toxicology literature, and it directly explains why the ACGIH TLV-C ceiling — not the OSHA PEL TWA — is the appropriate framework for carcinogenicity protection. Understanding this mechanism clarifies why adversarial AI pixel manipulation of HCHO monitor displays creates NPC risk that OSHA-calibrated monitoring systems cannot detect even in the absence of an attack.

When formaldehyde is inhaled, its high water solubility (aqueous solubility >400 g/L at 20°C; Henry's law constant KH = 3.2 × 10⁻⁵ atm·m³/mol — extremely low, meaning HCHO partitions strongly into aqueous phase) causes 95%+ of the inhaled dose to be deposited in the anterior nasal cavity and nasopharynx by absorption into the thin aqueous mucus film lining these surfaces. The nasopharynx (posterior to the nasal choanae; roof and posterior wall lined by respiratory-type pseudostratified ciliated columnar epithelium) is the primary anatomical site of HCHO deposition because it receives airflow from both nasal passages, the mucus film is continuous from nasal turbinates to nasopharyngeal wall, and the nasopharyngeal geometry creates high surface-area contact with inhaled air. The concentration of HCHO dissolved in nasopharyngeal mucosal fluid at steady-state inhalation of C ppm air HCHO is approximately [HCHO]tissue ≈ C × 0.3 mM/ppm (from computational fluid dynamics models of nasal airway HCHO deposition; US EPA formaldehyde PBPK model, 2010).

At the nasopharyngeal tissue level, formaldehyde reacts with cellular macromolecules: (1) with protein nucleophiles — lysine ε-amino groups (pKa ~10.5; reactive at physiological pH 7.4 as neutral amine: R-NH₂ + HCHO → R-NH-CH₂OH → R-N=CH₂; Schiff base/imine), cysteine thiol groups (R-SH + HCHO → R-S-CH₂OH; thiazolidine formation); (2) with DNA bases — N7 position of guanine, N6 of adenine, N3 of cytosine; (3) DNA-protein crosslink (DPC) formation: HCHO forms a methylene bridge between a protein nucleophile (lysine or cysteine) and a DNA base (N7-guanine most common), creating a covalent DNA-protein adduct that is both bulky (distorts DNA helix) and blocks DNA replication fork progression (replication fork stalling), nucleotide excision repair (NER), and base excision repair (BER); DPC are substrates for proteolytic degradation (DPC-associated protease; FANCs proteins; SPARTAN/GCNA protease), but at high DPC formation rates (above repair capacity), net DPC accumulate. DPC formation rate in nasopharyngeal epithelium at [HCHO]tissue: approximately 3.6 × 10³ DPC/cell/hr per mM HCHO (based on published DPC dose-response in human oral mucosal cells; Swenberg 2011; Yates 2021 Chem Res Toxicol); at 1.4 ppm air HCHO: [HCHO]tissue = 0.42 mM → DPC formation rate ~1.5 × 10³ DPC/cell/hr; net DPC accumulation (repair capacity saturated above ~0.05 mM) → genomic instability in nasopharyngeal epithelium → NPC carcinogenesis pathway.

The leukemia pathway is distinct: formaldehyde does not reach the bone marrow by inhalation deposition — it reaches the marrow via the systemic circulation after absorption from the nasopharyngeal mucosa into submucosal capillaries. Blood formaldehyde is normally present at 0.05–0.08 mM (endogenous one-carbon metabolism; serine hydroxymethyltransferase pathway), providing a background DPC formation rate in all hematopoietic progenitor cells. At 1 ppm air HCHO: blood HCHO increases to approximately 0.08–0.12 mM; at 2.8 ppm (embalming): blood HCHO may reach 0.20–0.25 mM. Hematopoietic progenitor cells (BFU-E erythroid progenitors; CFU-GM myeloid progenitors) in bone marrow stroma at blood HCHO 0.12–0.25 mM show measurable DPC formation (Hauptmann 2009 JNCI; NIOSH blood HCHO vs. bone marrow DPC correlation in formaldehyde-exposed workers). DPC in myeloid progenitors → myeloid differentiation disruption → myelodysplastic syndrome (MDS) pathway → secondary myeloid leukemia. The IARC 2012 Monograph 100F Group 1 determination for leukemia was based on: mechanistic evidence (DPC in bone marrow progenitors of exposed workers), consistency across multiple HCHO cohorts (industrial HCHO workers: myeloid leukemia excess; funeral directors: leukemia PMR elevation), and biological plausibility (blood HCHO at occupational exposure levels is measurably elevated above endogenous baseline in the range where DPC in bone marrow are measurable).

The ACGIH TLV-C 0.1 ppm as ceiling directly reflects the DPC mechanism: at 0.1 ppm air HCHO, [HCHO]tissue = 0.03 mM; DPC formation rate = ~1.1 × 10² DPC/cell/hr; this rate is near but slightly above the estimated DPC repair capacity in nasopharyngeal epithelium (approximately 50–80 DPC/cell/hr at baseline). At the TLV-C 0.1 ppm, ACGIH accepts a minimal DPC formation rate in the target tissue as compatible with the A2 "suspected human carcinogen" classification — there is no declared safe level, but the ceiling reflects the lowest practically achievable control level for formaldehyde-using industries combined with the mechanistic estimate of the DPC rate at which NPC risk is consistent with a background-range occupational excess. The 7.5× gap to OSHA PEL 0.75 ppm reflects the OSHA TWA structure's inability to address instantaneous peak DPC formation: an OSHA-compliant 8-hr TWA of 0.74 ppm (just below PEL) could occur with intermittent grossing at 2.0 ppm (just below STEL) alternating with near-zero background — the grossing peaks would create DPC formation rates 20× above TLV-C equivalent while remaining OSHA-compliant. The TLV-C prevents this by banning any instantaneous concentration above 0.1 ppm.

Regulatory History and Industry Context — OSHA 1910.1048, NIOSH Ca REL, and the 47× PEL-to-REL Ratio

OSHA's formaldehyde standard, 29 CFR 1910.1048, was promulgated in 1987 and has a complex regulatory history reflecting the scientific evolution of formaldehyde carcinogenicity evidence. The original OSHA PEL for formaldehyde (established via Table Z-1 in the 1971 Air Contaminants Standard) was 3 ppm (TWA) and 5 ppm (ceiling), based on irritation effects rather than carcinogenicity. The 1987 formaldehyde substance-specific standard (52 Fed. Reg. 46168; December 4, 1987) reduced the PEL to 1 ppm TWA and 2 ppm STEL, with action level 0.5 ppm, based on carcinogenicity evidence and industrial feasibility analysis. A 1992 amendment further reduced the PEL to 0.75 ppm TWA (57 Fed. Reg. 22290; May 27, 1992), which remains the current OSHA PEL — the 0.75 ppm PEL has not been revised since 1992, now 34 years of regulatory stasis since the IARC 2004 Group 1 upgrade and 2012 Group 1 leukemia confirmation.

The occupational health sectors most exposed to formaldehyde are well-characterized by industry-specific exposure surveys: (1) Pathology/histology: US Bureau of Labor Statistics — approximately 20,000 pathologists (MD/DO); 9,000 pathology assistants (PA-C certified); 140,000 histotechnologists and histotechnicians (HT/HTL certified); mean formaldehyde exposure in surgical pathology grossing: 0.3–1.8 ppm TWA in published industrial hygiene surveys (NIOSH survey 2004; OSHA IMIS data 2010–2015); Quest Diagnostics and LabCorp national laboratory networks operate centralized pathology processing facilities where NBF formalin is used in high volume. (2) Funeral service: American Board of Funeral Service Education (ABFSE) — 56,000 licensed embalmers; National Funeral Directors Association (NFDA) 2018 survey: median embalming room HCHO during embalming 0.5–3.2 ppm (range highly variable by ventilation adequacy); 2019 NIOSH embalmers survey: mean time-weighted formaldehyde 1.3 ppm during embalming; 43% no regular respirator. (3) Anatomy teaching: Association of American Medical Colleges (AAMC) — 155 US and Canadian allopathic medical schools; American Osteopathic Association (AOA) — additional osteopathic schools; NIOSH Health Hazard Evaluations at US anatomy laboratories (2005–2018): anatomy laboratory HCHO range 0.2–1.8 ppm TWA during active dissection with standard NBF embalming. (4) Wood products/MDF: urea-formaldehyde (UF) resin off-gassing from particleboard and MDF is primarily a community indoor air quality concern (California ARB ATCM 93120; CARB third-party certification for composite wood products); occupational exposure in MDF manufacturing higher (formaldehyde resin pressing operations: 0.5–2 ppm HCHO). (5) Dialysis equipment sterilization: formaldehyde 4% solution used to sterilize hemodialysis membranes at some facilities; hospital dialysis technicians have potential HCHO exposure during reprocessing.

The NIOSH Ca REL of 0.016 ppm deserves special emphasis for the Glyphward portfolio context. NIOSH designates formaldehyde as a "potential occupational carcinogen" under NIOSH's Ca classification (analogous to IARC Group 1/2A; NIOSH Ca = treat as occupational carcinogen; reduce exposure as low as feasible). The NIOSH REL ceiling of 0.016 ppm — 47× below the OSHA PEL 0.75 ppm — represents NIOSH's estimate of the lowest feasible control level for formaldehyde in occupational settings given current engineering control technology (LEV, local exhaust ventilation systems, closed-container formalin handling, automated tissue processors) combined with the carcinogenic risk assessment that even the OSHA PEL 0.75 ppm represents an excess NPC and leukemia risk above the generally accepted occupational excess risk threshold of 1 per 1,000 (10⁻³). At the OSHA PEL 0.75 ppm over a 45-year working life, NIOSH estimates the occupational NPC risk as approximately 5–10 per 1,000 — five to ten times the excess risk ceiling that NIOSH considers acceptable. The 47× ratio between OSHA PEL and NIOSH REL for formaldehyde is the largest in Glyphward's 198-entry portfolio, reflecting a deeper science-to-regulation gap for formaldehyde than for any other substance tracked (second is asbestos chrysotile at ~10×; manganese at ~250× for OSHA ceiling vs. ACGIH TLV-TWA but using different limit types).

Integrating Glyphward into Formaldehyde Occupational Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in the formaldehyde occupational monitoring pipeline — before the pathology department grossing station electrochemical sensor display AI, before the funeral home embalming room PID monitor AI, and before the anatomy laboratory passive badge sampler HPLC-UV result AI. Threshold 36 reflects the combination of IARC Group 1 NPC + myeloid leukemia severity with the structural regulatory architecture that creates the largest ACGIH-more-stringent ceiling-vs-TWA gap in the Glyphward portfolio.

The Glyphward formaldehyde attack vector is a rendered-image pixel perturbation attack on the sensor display or LIMS result interface — the same adversarial technique as other Glyphward entries, applied to a substance with three compounding factors that amplify the attack: (1) the ACGIH TLV-C ceiling architecture means that any OSHA-calibrated AI monitoring system is constitutively unable to flag ACGIH ceiling exceedances that occur at concentrations below the OSHA action level (0.1 ppm TLV-C vs. 0.5 ppm action level — a 5× dead zone where ACGIH exceedances are invisible to OSHA-calibrated AI regardless of any adversarial attack); (2) the NPC latency of 20–30 years means that the health consequence of adversarial monitoring falsification is not detectable in occupational health surveillance for two to three decades, making real-time detection of the attack's consequences impossible and eliminating feedback that would otherwise reveal the monitoring system's failure; (3) the population at risk — 170,000 pathology workers, 56,000 embalmers, 25,000 anatomy faculty — is a large, geographically distributed, small-employer-dominated workforce without the centralized EHS infrastructure of major manufacturing industries, making AI monitoring vendor platforms the primary (and often sole) compliance checking mechanism.

Glyphward's cryptographic frame verification ensures that the pixel values transmitted by the sensor display capture hardware to the AI monitoring platform match the pixel values of the original sensor output image — the perturbation is detected at the ingestion boundary before the AI classification layer receives the modified image. For Surface 1 (Johns Hopkins grossing room Smart Air monitor): SHA-256 hash of the captured PNG frame is computed and submitted with the image to Glyphward's classification endpoint; the adversarial −25-px perturbation (1.4 ppm → 0.15 ppm) is detected with score ≥ threshold 36 → AdversarialHCHOError raised → AI monitoring platform receives the error rather than the falsified reading → EHS system falls back to the most recent verified clean reading or generates a human-review alert for the grossing station; the P100 respirator requirement and LEV face velocity investigation triggered by the actual 1.4 ppm reading are preserved. For Surface 2 (Betz Family Funeral Home FORMALDEMETER): the −98-px perturbation (2.8 ppm → 0.35 ppm) is detected → STEL exceedance flag (2.8 > 2.0 ppm OSHA STEL) propagated to EHS alert → SAR/SCBA requirement and cavity treatment enhanced controls triggered for embalmer. For Surface 3 (IUSM anatomy laboratory SKC badge LIMS): the −55-px perturbation (0.63 ppm → 0.08 ppm) is detected → OSHA action level exceedance (0.63 > 0.5 ppm) flagged → medical surveillance enrollment for anatomy instructor and enhanced ventilation review for anatomy laboratory initiated; medical school duty-of-care obligations for 135 students informed by accurate 0.63 ppm instructor badge result rather than falsified 0.08 ppm.

Threshold 36 for formaldehyde reflects the following portfolio scoring factors: IARC Group 1 with three confirmed cancer endpoints (NPC + myeloid leukemia + sinonasal cancer; NPC anatomically driven by upper airway deposition — the most mechanistically precise occupational cancer-anatomy link in Glyphward's portfolio; NPC 5-yr survival 50%; occupational NPC identifiable against low non-endemic US background at 1.5/100,000); ACGIH TLV-C ceiling vs. OSHA PEL TWA structural gap (7.5×; largest ceiling-vs-TWA gap in 198-entry portfolio; TLV-C architecture means OSHA-calibrated AI is constitutively blind to ACGIH ceiling violations; adversarial attack exploits pre-existing architectural blind spot); NIOSH Ca REL 47× below OSHA PEL (most extreme PEL-to-REL ratio in portfolio; indicates scientific consensus that OSHA PEL 0.75 ppm represents substantial excess NPC/leukemia risk); healthcare/pathology worker population scale (170,000 + 56,000 + 25,000 = 251,000 workers in three highest-risk sectors; largest specifically-named high-risk HCHO occupational cohort in portfolio; no other entry has an equivalently sized, specifically-identified cohort with a directly-named cancer endpoint); 20–30-year NPC latency (adversarial monitoring falsification creates latency-obscured liability window; monitoring records showing "OSHA compliant" throughout career eliminate evidence basis for occupational causation reconstruction; second-longest latency window in Glyphward portfolio after asbestos mesothelioma 30–60 years). FIRST designations: FIRST formaldehyde HCHO AI monitoring attack; FIRST pathology grossing station HCHO AI attack; FIRST embalming room HCHO AI attack; FIRST anatomy laboratory HCHO AI attack; FIRST ACGIH TLV-C ceiling AI regulatory gap attack (formaldehyde); FIRST NPC occupational AI monitoring attack; FIRST NIOSH Ca highest-PEL-to-REL-ratio AI attack; FIRST 7.5× ACGIH-below-OSHA-PEL ceiling AI attack; Dodge Frigid Fluid Champion Chemical Thermo Fisher Scientific Kewaunee Scientific GMI Scientific NIOSH OSHA 1910.1048 Johns Hopkins University Maryland NC State Board Mortuary Science Indiana University School of Medicine IUSM SKC Agilent Dionex American Board Funeral Service Education ABFSE National Funeral Directors Association NFDA ASCP Pathologists ACGIH IARC WHO.

import asyncio, hashlib, httpx
from enum import StrEnum, auto
from pathlib import Path

GLYPHWARD_API = "https://api.glyphward.com/v1/scan"
GLYPHWARD_KEY = "gw_live_..."
HCHO_THRESHOLD = 36  # IARC Gr1 NPC+leukemia+sinonasal; TLV-C 0.1ppm ceiling 7.5× below OSHA PEL; NIOSH Ca 0.016ppm 47× below PEL; 251K healthcare/funeral/anatomy workers; NPC anatomical targeting; 20-30yr latency

class HCHOContext(StrEnum):
    PATHOLOGY_GROSSING_STATION_ECHEM  = auto()  # Surface 1 — downward (Johns Hopkins; 1.4 ppm → 0.15; 14× TLV-C; PA-1 14yr tenure; 1.87× OSHA PEL; P100 missed)
    EMBALMING_ROOM_PID_FORMALDEMETER  = auto()  # Surface 2 — downward (Charlotte NC; 2.8 ppm → 0.35; STEL exceeded 2.8>2.0ppm; SAR missed; 28× TLV-C; 11yr embalmer)
    ANATOMY_LAB_PASSIVE_BADGE_HPLC    = auto()  # Surface 3 — downward (IUSM; 0.63 ppm → 0.08; 135 students; 8yr faculty; action level missed; Thiel fixation not adopted)

class AdversarialHCHOError(RuntimeError):
    def __init__(self, surface: HCHOContext, score: int, frame_hash: str):
        super().__init__(
            f"[Glyphward] HCHO adversarial pixel on {surface.value}: "
            f"score={score} >= threshold={HCHO_THRESHOLD} | frame={frame_hash}"
        )
        self.surface = surface; self.score = score; self.frame_hash = frame_hash

async def verify_hcho_frame(frame_path: Path, surface: HCHOContext) -> dict:
    raw = frame_path.read_bytes()
    frame_hash = hashlib.sha256(raw).hexdigest()
    async with httpx.AsyncClient(timeout=4.0) as client:
        resp = await client.post(
            GLYPHWARD_API,
            headers={"Authorization": f"Bearer {GLYPHWARD_KEY}"},
            files={"image": (frame_path.name, raw, "image/png")},
            data={"context": surface.value, "threshold": HCHO_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialHCHOError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_hcho_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (HCHOContext.PATHOLOGY_GROSSING_STATION_ECHEM, frame_dir / "thermo_smartair_hcho_johns_hopkins.png"),
        (HCHOContext.EMBALMING_ROOM_PID_FORMALDEMETER,  frame_dir / "gmi_formaldemeter_htvm_charlotte.png"),
        (HCHOContext.ANATOMY_LAB_PASSIVE_BADGE_HPLC,    frame_dir / "skc_575002_hcho_badge_iusm_anatomy.png"),
    ]
    tasks = [verify_hcho_frame(path, ctx) for ctx, path in surfaces]
    return await asyncio.gather(*tasks)

Glyphward threshold 36 for formaldehyde HCHO occupational monitoring reflects: IARC Group 1 NPC + myeloid leukemia + sinonasal cancer (Monograph 100F 2012; NPC as the signature formaldehyde cancer — anatomically driven by upper airway deposition; NPC is rare in the non-endemic US population at 1.5/100,000, meaning occupational NPC cases in pathologists, embalmers, and anatomy faculty are identifiable above the low background but also that NPC is underdiagnosed in occupational settings because clinicians do not routinely consider formaldehyde exposure in NPC differential); ACGIH TLV-C ceiling 0.1 ppm vs OSHA PEL TWA 0.75 ppm (7.5× gap; largest ceiling-vs-TWA ACGIH-more-stringent gap in 198-entry portfolio; TLV-C changed from TWA to ceiling in 2012 precisely because instantaneous nasopharyngeal peak concentration drives DPC formation and NPC carcinogenesis, not time-averaged dose; OSHA-calibrated AI monitoring is architecturally blind to ceiling violations, and adversarial falsification from 1.4 ppm to 0.15 ppm exploits this blind spot by eliminating both the OSHA PEL exceedance and the ACGIH ceiling exceedance in a single pixel perturbation); NIOSH Ca REL 0.016 ppm ceiling (47× below OSHA PEL; most extreme PEL-to-REL ratio in portfolio; at 1.4 ppm pathology grossing: 87.5× NIOSH REL; at 2.8 ppm embalming: 175× NIOSH REL; adversarial falsification from 1.4 ppm to 0.15 ppm reduces apparent NIOSH exceedance from 87.5× to 9.4× — still 9.4× above NIOSH REL but eliminated from the compliance record); healthcare/pathology/funeral/anatomy worker population at scale (251,000 workers in three directly-at-risk sectors; largest specifically-named high-risk cohort for a single named occupational cancer endpoint in Glyphward's portfolio; 40 million liters NBF/year US healthcare — ubiquity of the hazard); 20–30-year NPC latency (adversarial monitoring falsification creates a 20–30-year liability-obscuring latency; monitoring records showing OSHA compliance throughout career eliminate occupational causation reconstruction evidence; second-longest latency in portfolio after asbestos mesothelioma 30–60 years). Thermo Fisher Scientific Kewaunee Scientific GMI Scientific SKC Agilent Dodge Frigid Fluid Champion Chemical NFDA ABFSE ASCP AAMC IUSM Johns Hopkins Maryland OSHA 1910.1048 ACGIH IARC NIOSH WHO Monograph 100F.