CoChat
Sign Up
Cited by 10Information and Cyber SecurityPrivacy, Security, and Data ProtectionEthics and Social Impacts of AI

Digital shadow AI risk theory (DART): A framework for managing data disclosure and privacy risks of AI tools at work

Open access
Sebastian, Glorin·2026·Technological Forecasting and Social Change
DOI 10.1016/j.techfore.2026.124697
Open this paper in CoChat — free account

Abstract

The swift adoption of generative and agentic AI tools in workplace settings has introduced new organizational risks related to data disclosure, privacy, and governance. This study introduces the Digital Shadow AI Risk Framework (DART), which identifies and explains the behavioral and organizational risks arising from the informal and often unregulated use of AI tools by employees. DART comprises six interrelated risk dimensions: unintentional disclosure risk, the trust–dependence paradox, data sovereignty conflict, knowledge dilution, the ethical black box problem, and organizational feedback loops. The framework is evaluated using a three-wave survey research combining hypothesis testing and covariance-based structural equation modeling (CB-SEM) across three cross-industry surveys of professionals (Survey-1: N = 374; Survey-2: N = 179; Survey-3: N = 220). Survey-3 introduced multi-item latent measures enabling direct tests of H4, H6, and the full H8 mediation chain (opacity → trust → comfort → disclosure). Results support six of eight hypotheses. Knowledge dilution is confirmed through replication, while data sovereignty conflict consistently operates as a boundary condition rather than a direct predictor. The findings reveal persistent gaps in employee awareness, training, and organizational controls surrounding AI use. DART's contribution lies in distinguishing Shadow AI from traditional Shadow IT by showing how everyday, efficiency-driven AI use embeds risk into routine k…

Integrity signals

Registry

Crossref · 10.1016/j.techfore.2026.124697

Cites

19 of 19 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Cites

Flagged first, then by citations

Research works printed but not linked: 6

  1. Amershi, 2022, Six human-centered artificial intelligence grand challenges, ACM Trans. Interact. Intell. Syst., 12
  2. Dwivedi, 2023, Artificial intelligence (AI): multidisciplinary perspectives on emerging challenges, opportunities, and agenda for research, practice and policy, Int. J. Inf. Manag., 71
  3. Gandhi, 2025, Approaches to responsible governance of GenAI in organizations, arXiv preprint arXiv:2504.17044. · arXiv
  4. Kosmyna, N., Hauptmann, E., Yuan, Y.T., Situ, J., Liao, X.-H., Beresnitzky, A.V., Braunstein, I., Maes, P., 2025. Your brain on ChatGPT: accumulation of cognitive debt when using an AI assistant for essay writing task. arXiv preprint arXiv:2506.08872. https://arxiv.org/abs/2506.08872. · arXiv
  5. Raji, 2019, Actionable auditing: investigating the impact of public AI deployments. proceedings of the AAAI/ACM conference on AI, ethics, and society, Acad. Med., 1, 429
  6. Ryan, 2021, The ethics of artificial intelligence: mapping the debate, Philos. Technol., 34, 1023

Cited by

Sorted by citations

Related

Authors
Sebastian, G.
Affiliations
Université de Bordeaux (FR)
Corresponding author
Sebastian, Glorin
ORCID
Sebastian, Glorin
Venue
Technological Forecasting and Social Change
Publisher
Elsevier BV
Published
2026-08
Volume
229
Pages
124697
ISSN
0040-1625 (ISSN-L)
DOI
10.1016/j.techfore.2026.124697
OpenAlex
W7155494220
Cited by
10
References
19
Citations by year
2026: 10
FWCI
45.44
Type
journal article
Language
en
Open access
Licence
Locations
Technological Forecasting and Social Change (publishedVersion, cc-by)
Funders
Google
Topics
  • Physical Sciences › Computer Science › Information Systems › Information and Cyber Security
  • Social Sciences › Sociology and Political Science › Privacy, Security, and Data Protection
  • Social Sciences › Safety Research › Ethics and Social Impacts of AI
Keywords
shadow AI, generative AI in the workplace, agentic AI, privacy risks, AI governance, data sovereignty, AI opacity, organizational knowledge, responsible AI use

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review