The covert use of unauthorised AI tools at work ("Shadow AI") is now widely documented: recent global surveys report that a majority of employees conceal their AI use from employers (Gillespie et al., 2025: 57% across 48,340 respondents in 47 countries; WalkMe/SAP, 2025: 49% hiding use to avoid judgement; Ivanti, 2025: 32% hiding use for fear of penalty). This literature establishes the prevalence and motives of concealment, but treats it as a governance, security, or attitudinal problem. What remains unaddressed is the occupational health dimension: the act of concealment itself imposes a sustained cognitive cost on the worker, yet no workload instrument captures it. This contribution introduces Concealment Load — a novel psychosocial workload construct — and operationalises it as a measurable dimension within an extended NASA-TLX instrument (NASA-TLX-9), alongside two further AI-mediated dimensions: AI Trust and Override Anxiety. To the author's knowledge, no prior instrument measures the cognitive cost of concealment as an occupational workload dimension. The construct is grounded in the Job Demand-Control model and the COPSOQ psychosocial tradition, and anchored in Directive 89/391/EEC (Art. 6(2)(d)), the EU AI Act (Arts. 14 and 26(7)), and NIS2 (Art. 21(2)(g)). A practitioner field observation across 72 Romanian SME risk assessments found that none addressed AI-related cognitive or psychosocial load, evidencing a systemic measurement gap. We present the construct, its theoretical foundations, and a pilot validation protocol, arguing that Concealment Load offers OSH practitioners an auditable means of rendering an invisible, highly prevalent workplace risk measurable.
Daniel Vicentiu (Tue,) studied this question.