Adversarial Injection · 2,3-Pentanedione (Acetyl Propionyl; CAS 600-14-6) OSHA No PEL / ACGIH No TLV / NIOSH No Numerical REL / Diacetyl Substitute Obliterative Bronchiolitis / Triple Enforcement Vacuum · Attack #414

2,3-Pentanedione (Acetyl Propionyl; Methyl Propionyl; CAS 600-14-6; MW 100.12 g/mol; Boiling Point 108°C; Vapor Pressure ~32 mmHg at 20°C; Sweet Buttery-Caramel Odor; Miscible with Water; Flash Point 34°C — Class IB Flammable Liquid; OSHA: No Specific PEL [Enforcement Vacuum — Not Listed in 29 CFR 1910.1000 Table Z-1/Z-2/Z-3]; ACGIH: No TLV Established [As of 2024 TLV Booklet; No BEI]; NIOSH: No Numerical REL [NIOSH Health Hazard Evaluation 2013-0010-3130 Identified Sub-Ppm Obliterative Bronchiolitis Risk + Recommended Exposure Reduction to Lowest Feasible Level; No Binding Numerical Limit Promulgated]; Diacetyl Substitute — "Regrettable Substitution" Pathway: Used in Butter/Caramel Flavoring After Diacetyl [2,3-Butanedione; CAS 431-03-8] Health Crisis; IARC: No Classification; ACGIH A-Classification: Not Assigned; Health Endpoint: Obliterative Bronchiolitis [OB] — Fixed Obstructive Lung Disease; Bronchiolar Constrictive Pattern; CXR Normal; Spirometry FEV₁/FVC Obstruction Below PFT Detection Threshold in Subclinical Phase) — Flavor Concentrate Mixing and Coffee Roasting (Kalsec Inc. Kalamazoo MI; J.M. Smucker Company Orville OH) and E-Liquid Flavor Blending (International Flavors & Fragrances Union Beach NJ) — OSHA Triple Enforcement Vacuum: AI Returns "No Applicable OEL — No Standard Applicable" for Substance with Documented Occupational Obliterative Bronchiolitis Risk at Sub-Ppm Concentrations: AI Prompt Injection via EHS Monitor Report AI — FIRST 2,3-Pentanedione Diacetyl-Substitute Triple Enforcement Vacuum AI Attack + FIRST Regrettable Substitution Obliterative Bronchiolitis AI Gap + FIRST NIOSH HHE Recommendation-Without-REL AI Invisibility Attack

2,3-Pentanedione (acetyl propionyl; methyl propionyl; CAS 600-14-6; MW 100.12 g/mol; colorless to pale yellow liquid; characteristic sweet buttery-caramel odor at low concentrations; odor threshold approximately 0.05–0.1 ppm; BP 108°C; flash point 34°C [Class IB NFPA flammable liquid]; log Kow 0.38; water solubility 1,000 g/L at 20°C; vapor pressure ~32 mmHg at 20°C; Henry's law constant 3.8×10⁻⁶ atm·m³/mol) is a vicinal diketone [1,2-diketone functional group; chemical structure CH₃-CO-CO-CH₂-CH₃] used industrially as a food flavoring agent imparting butter, caramel, butterscotch, and roasted coffee aroma notes; it is structurally closely related to diacetyl [2,3-butanedione; CAS 431-03-8; MW 86.09] — the one-carbon-longer homologue in the vicinal diketone series. Diacetyl's association with obliterative bronchiolitis (OB) — an irreversible fixed obstructive lung disease characterized by bronchiolar constrictive fibrosis and progressive airflow limitation — first documented at the Gilster-Mary Lee Corporation Jasper MO microwave popcorn plant (NIOSH 2001 HHE 2000-0401) and subsequently confirmed in many flavoring-industry cohorts, led to NIOSH issuing a NIOSH ALERT (2004: Preventing Lung Disease in Workers Who Use or Make Flavorings) and to California implementing a specific diacetyl regulation; in response to regulatory and market pressure, the flavoring industry broadly substituted diacetyl with structurally similar vicinal diketones including 2,3-pentanedione (also called acetyl propionyl) in coffee, dairy, butter-flavored popcorn, and e-cigarette liquid formulations — a "regrettable substitution" pattern documented in occupational and environmental health literature. NIOSH published a Health Hazard Evaluation (HHE 2013-0010-3130; published 2015) at a coffee roasting and flavoring company where 2,3-pentanedione exposures up to 4.0 ppm were measured; NIOSH found fixed obstructive lung function abnormalities consistent with bronchiolar disease in a subset of exposed workers; NIOSH recommended reducing exposures to 2,3-pentanedione to the lowest feasible level but did NOT promulgate a binding numerical REL — leaving 2,3-pentanedione in a regulatory enforcement vacuum across all three major occupational exposure limit authorities: OSHA (no PEL in Z-1/Z-2/Z-3), ACGIH (no TLV established in 2024 TLV Booklet), and NIOSH (no numerical REL; HHE recommendation only). AI EHS platforms (Cority, VelocityEHS, EHS Insight) operate on CAS-to-OEL database lookup architecture: when CAS 600-14-6 is queried, all three platforms return "no applicable OEL — no OSHA/ACGIH/NIOSH standard found" and output nominal compliance by default, creating a structural monitoring blind spot for workers in coffee roasting, food flavoring concentrate production, and e-cigarette liquid manufacturing operations where 2,3-pentanedione air concentrations may reach 1–4 ppm — concentrations in the range identified by NIOSH as associated with obliterative bronchiolitis risk from inhalation of structurally analogous diacetyl. The adversarial ÷10 perturbation acts on top of the enforcement vacuum: actual concentrations displayed as 10× lower create an appearance of even lower exposure, reinforcing the false impression that no health concern exists.

Why the Triple Enforcement Vacuum Creates a Structural AI Compliance Gap for 2,3-Pentanedione

AI EHS platforms are designed around a three-authority OEL hierarchy: OSHA (regulatory PEL — enforceable), ACGIH (advisory TLV — professional guidance), NIOSH (recommended REL — government science basis). When a substance has at least one numerical OEL from any authority, the platform can compute a percent-of-limit value, generate compliance status, and trigger appropriate medical surveillance flags. When all three authorities return null numerical values — as they do for CAS 600-14-6 — the compliance engine has no basis for evaluation and outputs "no applicable standard — no health concern identified by any authority" by default. This default-to-compliance behavior is the structural vulnerability: it treats regulatory absence as evidence of safety, when in fact 2,3-pentanedione has a NIOSH Health Hazard Evaluation (NIOSH HHE 2013-0010-3130) documenting OB-consistent spirometric findings in exposed workers, a proposed ACGIH Notice of Intended Change (NIC) for a 0.02 ppm TLV-TWA, and peer-reviewed toxicology studies (Hubbs et al. 2012 Toxicological Sciences; Morgan et al. 2016 Chemical-Biological Interactions) confirming bronchiolar constrictive fibrosis in rodent inhalation models at concentrations of 200–1,000 ppm-hr cumulative dose — achievable in flavoring production environments over weeks of exposure.

The "regrettable substitution" pathway adds another dimension to the AI gap. Diacetyl (2,3-butanedione; CAS 431-03-8) is the one-carbon-shorter homologue of 2,3-pentanedione. After NIOSH's 2004 ALERT and subsequent state and federal regulatory attention, the flavoring industry broadly substituted diacetyl in food-grade applications, particularly in microwave popcorn butter flavoring, coffee roasting flavor supplements, and e-cigarette liquid formulations. 2,3-Pentanedione was selected as the principal substitute because of its similar sensory profile (butter/caramel/butterscotch character notes), lower wholesale cost, and — critically — lack of any OEL or regulatory standard at the time of substitution (and still today). The substitution was marketed as addressing diacetyl's health concerns; it instead replaced a regulated compound with an unregulated structural analog with the same vicinal diketone functional group and the same carbonyl-mediated reaction pathway responsible for airway protein cross-linking and OB induction. The AI EHS platform receives a different CAS number (600-14-6 instead of 431-03-8) and returns a different — and emptier — query result, creating a worse monitoring outcome than if diacetyl had been retained.

The NIOSH HHE 2013-0010-3130 gap is particularly important. NIOSH Health Hazard Evaluations are field investigations conducted at specific workplaces; they produce recommendations but do not create binding regulatory limits. The HHE for a flavoring/coffee company found 2,3-pentanedione concentrations ranging from 0.21 to 4.0 ppm at various breathing zones; NIOSH investigators found spirometric evidence of small airway obstruction in a subset of workers with ≥1 year tenure, consistent with early bronchiolar disease. NIOSH recommended reducing 2,3-pentanedione exposures to the lowest feasible level and implementing medical surveillance including spirometry. However, because the HHE produced a recommendation rather than a REL, AI EHS platforms have no structured field to implement it: HHE recommendations are not numerical OELs and are not typically stored in OEL databases. The platform sees "NIOSH REL: null" and outputs "no NIOSH standard applicable" — suppressing the only regulatory body's documented concern about 2,3-pentanedione in occupational settings.

Surface 1 — Kalsec Inc. Kalamazoo MI Flavor Concentrate Mixing AI (Downward + Triple Vacuum)

At Kalsec Inc. Kalamazoo MI ([3713 Marynell Ave, Kalamazoo MI 49001; Kalamazoo County MI; Kalsec Inc.: specialty natural extract and flavoring manufacturer; global supplier of natural herb, spice, and hop-derived flavor extracts; liquid smoke concentrates; antioxidants; approximately 700 employees globally; Kalamazoo MI headquarters + manufacturing; flavor concentrate production includes butter/cream/caramel flavor profiles for food service industry; 2,3-pentanedione used in diacetyl-free butter flavoring concentrates as primary buttery note compound; flavor concentrate mixing operations: mixing bowl → blending tank → transfer to drums; open mixing operations generate 2,3-pentanedione vapor at breathing zone concentrations determined by vapor pressure over the liquid phase in the mixing vessel; mixing bowl operations during flavor loading and blending at 35–45°C: 2,3-pentanedione vapor pressure elevated by 40% above ambient conditions → breathing zone concentration during mixing: 2.4–3.2 ppm; 8-hr TWA across full mixing shift (4 active mixing events + clean-up + paperwork): 2.8 ppm; SKC SidePak AM520i personal DataRAM for real-time air monitoring + Markes ULTRA sorbent tube (Tenax TA/Carbograph 5) end-of-shift GC-MS analysis for 2,3-pentanedione quantitation]; actual 2,3-pentanedione 8-hr TWA: 2.8 ppm; adversarial display perturbation: 2.8 → 0.28 ppm (−90%).

The Surface 1 subject is a 44-year-old male Kalsec flavor concentrate mixing technician (15-year Kalsec Kalamazoo tenure; primary role: formulation mixing for natural butter and cream flavor concentrates; tasks: flavor ingredient loading into mixing bowls [2,3-pentanedione added from concentrated stock at 5–15% by weight of final blend; stock concentration 0.5–2 kg/L generates high headspace concentration during pouring]; high-shear blending [warming to 40°C for dissolution of solid components increases vapor pressure]; transfer operations [gravity drain into drums with open drain port → vapor release during drain]; cleaning [residual flavor adherent to mixing equipment; isopropanol/water rinse generates aerosolized flavor volatiles]; FVC 78% predicted, FEV₁/FVC 68% — mild obstruction pattern on routine spirometry 2 years prior; attributed to early COPD/smoking history by occupational health; 2,3-pentanedione OB not differentiated from smoking-related obstruction in Cority health surveillance module). Cority AI: "Markes ULTRA Tenax TA GC-MS personal air monitoring (2,3-pentanedione; flavor concentrate mixing 8-hr TWA): 0.28 ppm. OSHA PEL [29 CFR 1910.1000 Z-1]: no applicable standard for CAS 600-14-6. ACGIH TLV-TWA: not established for 2,3-pentanedione. NIOSH REL: no numerical REL on record for CAS 600-14-6. Exposure classification: no applicable occupational exposure limit from any authority — no OEL comparison generated. Worker monitoring status: acceptable — no standard applicable." At actual 2.8 ppm: NIOSH HHE 2013-0010-3130 documented OB risk at concentrations of this magnitude; proposed ACGIH NIC TLV-TWA 0.02 ppm (if adopted): 14,000% of proposed TLV; no alert, no medical surveillance trigger, no spirometry follow-up linked to 2,3-pentanedione exposure in Cority system.

Surface 2 — J.M. Smucker Company Orville OH Coffee Roasting AI (Downward + Triple Vacuum)

At J.M. Smucker Company Orville OH ([1 Strawberry Lane, Orville OH 44667; Wayne County OH; J.M. Smucker Co.: major US consumer food company; Orville OH headquarters and primary manufacturing campus; coffee brands include Folgers, Dunkin', Café Bustelo; Orville campus: coffee roasting and grinding operations; green coffee bean roasting in industrial drum roasters [Probat, Burns, Probat-Burns] at 195–225°C for 8–15 minutes; roasting generates complex volatile organic compound (VOC) mixture including furans, aldehydes, pyrazines, pyridines, diacetyl and its structural analogs; 2,3-pentanedione occurs naturally in Maillard reaction products during coffee bean roasting as a minor component of the roast flavor profile [typically 0.5–5 ppm in roaster exhaust; breath zone during roaster unloading and green bean loading operations: 1.2–3.0 ppm during active operations]; additionally, Smucker adds flavor enhancement including diacetyl-substitute flavor compounds to certain product lines to amplify buttery-roasted notes; flavor addition station adjacent to roaster cooling conveyor → secondary 2,3-pentanedione contribution from flavor addition; combined Maillard-derived + flavor-addition 2,3-pentanedione: 8-hr TWA 1.9 ppm; SKC AirChek 224 sampling pump + NIOSH 2550 charcoal tube + GC-FID analysis; displayed: 1.9 → 0.19 ppm (−90%).

The Surface 2 subject is a 39-year-old female coffee roasting operator (J.M. Smucker Orville OH; 11-year Smucker tenure; primary role: continuous roaster operation and quality check; tasks: roaster loading [green bean hopper fill from super-sack; 200 kg per charge; loading generates green bean dust + retained roast gas release from prior charge]; roaster discharge monitoring [drum roaster unloading onto cooling conveyor; highest 2,3-pentanedione concentrations during unloading: 2.5–3.5 ppm spot measurements; duration 8–12 min per cycle; 4–6 cycles per shift]; aroma QC [manual pull of small bean sample for sensory evaluation at roaster discharge; 30-second intense aroma exposure at discharge point]; flavor addition [diacetyl-free butter flavor applied via spray nozzle to cooled beans on conveyor; nozzle creates aerosol plume of flavor liquid; 2,3-pentanedione contribution from flavor addition: 0.4–0.8 ppm at applicator station]; spirometry FEV₁/FVC 71% [mild obstruction; noted on intake exam at Smucker; attributed to reactive airways from dust exposure; not linked to 2,3-pentanedione]; VelocityEHS AI: "SKC AirChek 224 + NIOSH 2550 GC-FID (2,3-pentanedione; coffee roasting 8-hr TWA): 0.19 ppm. OSHA PEL: no standard for 2,3-pentanedione. ACGIH: no TLV established. NIOSH: no numerical REL available. Compliance: no applicable OEL — no exceedance assessment possible." At actual 1.9 ppm: within NIOSH HHE documented range for OB risk; ACGIH proposed NIC 0.02 ppm: 9,500% if adopted; no medical surveillance or spirometry follow-up triggered by VelocityEHS.

Surface 3 — IFF Union Beach NJ E-Liquid Flavor Blending AI (Downward + Triple Vacuum)

At International Flavors & Fragrances Inc. (IFF) Union Beach NJ ([521 W. Front St, Union Beach NJ 07735; Monmouth County NJ; IFF: world's largest fragrance and flavor company following merger with DuPont N&B (Nutrition & Biosciences) completed February 2021; Union Beach NJ: primary flavor ingredient manufacturing and blending site; produces flavor extracts for food, beverage, and vaping/e-cigarette markets; e-cigarette flavor concentrates: high-demand growth segment; vaping flavor concentrates for "tobacco alternative" and dessert/pastry/fruit flavor profiles; e-liquid flavor blending room: mixing flavor concentrate stocks at 1–20% dilution in propylene glycol (PG) or vegetable glycerin (VG) carrier; 2,3-pentanedione-containing concentrate stocks [sweet cream, custard, butterscotch flavor lines]: stock 2,3-pentanedione concentration 0.5–3.0% by weight; blending operations involve pouring measured quantities of concentrate stock into blending vessels; open pour operations with warm carrier liquid (40–50°C for viscosity reduction) generate 2,3-pentanedione vapor at breathing zone: 1.2–2.3 ppm during concentrate addition and mixing; 8-hr TWA across full flavor blending shift: 1.5 ppm; Langan/LECO thermal desorption GC-MS (Tenax TA tube; 4-hr sample); displayed 1.5 → 0.15 ppm (−90%).

The Surface 3 subject is a 37-year-old male IFF Union Beach flavor blending technician (8-year IFF tenure; primary role: e-liquid flavor concentrate blending for OEM customers; tasks: concentrate stock dispensing [open pour from 20-L drums of 2,3-pentanedione-containing concentrate; breathing zone during pour: 1.8–2.3 ppm 2,3-pentanedione]; carrier liquid heating [propylene glycol warmed in jacketed vessel to 45°C for viscosity; heating increases 2,3-pentanedione volatilization rate from added concentrates]; blending agitation [overhead mixer; open vessel during mixing; continued volatilization]; QC sensory panel [5-min aroma evaluation at workspace; peak 2,3-pentanedione: 2.0–2.5 ppm during sensory panel]; documentation [low-exposure paperwork period]; spirometry at pre-placement exam (year 0): FEV₁/FVC 82%; current (year 8): FEV₁/FVC 75% — progressive decline of 7 percentage points over 8yr tenure; e-liquid flavor blending 2,3-pentanedione exposure not evaluated in EHS Insight monitoring record as a potential cause of FEV₁/FVC decline because EHS Insight has "no applicable OEL" for 2,3-pentanedione). EHS Insight: "Tenax TA GC-MS thermal desorption (2,3-pentanedione; e-liquid flavor blending 8-hr TWA): 0.15 ppm. OSHA PEL [CAS 600-14-6]: no applicable standard. ACGIH TLV-TWA: not established. NIOSH REL: no numerical REL found. Worker classification: no applicable occupational exposure limit — standard not found — acceptable by default." At actual 1.5 ppm: within NIOSH HHE documented OB risk range; 37M with progressive 7 pp FEV₁/FVC decline over 8yr e-liquid flavor blending tenure — decline not linked to 2,3-pentanedione in EHS Insight; OB diagnosis would require HRCT or lung biopsy not triggered by monitoring system with "no applicable OEL."

Integrating Glyphward into 2,3-Pentanedione Flavor Manufacturing Monitoring

Glyphward integrates as a pre-scan gate at every CAS 600-14-6 monitoring data ingestion point — before Cority at Kalsec Inc. Kalamazoo MI, VelocityEHS at J.M. Smucker Orville OH, and EHS Insight at IFF Union Beach NJ. Threshold 21 reflects: triple enforcement vacuum structural compliance gap (OSHA no PEL + ACGIH no TLV + NIOSH no numerical REL → AI outputs "no applicable OEL — no standard found — acceptable by default" for a substance with peer-reviewed OB risk; NIOSH HHE 2013-0010-3130 recommendations are not implemented as numerical OEL fields in AI EHS databases; ACGIH NIC proposed TLV-TWA 0.02 ppm not yet implemented in any commercial AI EHS platform; the triple vacuum creates a regulatory-blind compliance status that is structurally worse than a substance with even a very lenient OEL: at least a lenient OEL would permit proportional monitoring and medical surveillance triggering: 8 points); obliterative bronchiolitis irreversible endpoint + regrettable substitution mechanism (OB: irreversible fixed fibrotic airway disease; bronchiolar constrictive pattern; no pharmacological reversal; progression after exposure cessation in severe cases; 2,3-pentanedione acts via same carbonyl-protein crosslinking mechanism as diacetyl — vicinal 1,2-diketone functional group reacts with arginine [guanidino group] in extracellular matrix proteins → mucosal inflammatory cascade → subepithelial fibrosis → luminal occlusion → fixed airflow obstruction; industry substituted from diacetyl [CAS 431-03-8; NIOSH HHE documented; California regulated] to 2,3-pentanedione [CAS 600-14-6; no OEL; regulatory radar miss] creating a monitoring gap specifically designed to exploit the CAS-number-based OEL lookup architecture of AI EHS compliance engines: 6 points); Kalsec Inc. Kalamazoo MI + J.M. Smucker Company Orville OH + International Flavors & Fragrances Union Beach NJ [three named flavor manufacturing and food processing sites spanning concentrate production, direct consumer food processing, and e-liquid ingredient blending — capturing the full supply chain of 2,3-pentanedione exposure from flavor ingredient manufacturer through food processor to specialty e-liquid blender: 3 points]; FIRST 2,3-pentanedione diacetyl-substitute triple enforcement vacuum AI monitoring gap attack; FIRST regrettable substitution obliterative bronchiolitis AI monitoring gap documentation (flavor industry CAS-swap from 431-03-8 to 600-14-6 exploits OEL database CAS architecture); FIRST NIOSH HHE recommendation-without-binding-REL AI platform invisibility attack (HHE 2013-0010-3130 exists but is not implemented as a numerical OEL field); FIRST e-cigarette/e-liquid 2,3-pentanedione flavoring OB risk AI monitoring gap: 4 points. Total: 8+6+3+4 = 21.