Prompt 9 — Explainable AI for Government Decision Support
Conduct a comprehensive review of explainable AI, trusted AI, AI governance, provenance systems, traceability frameworks, assurance cases, auditability, and decision transparency. Evaluate how government and defense organizations can deploy AI decision-support systems while maintaining accountability, explainability, and regulatory compliance.
Generated on: June 07, 2026
Now I have gathered comprehensive research across all key topics. Let me create a final literature review with proper structure and citations. Based on the extensive search results, I&#39;ll now write the comprehensive literature review.
Comprehensive Literature Review: Explainable AI for Government Decision Support
1. Introduction and Foundational Concepts
The deployment of artificial intelligence in government decision-support systems represents one of the most consequential technological shifts in public administration. As governments globally adopt AI for critical functions ranging from welfare eligibility determinations to defense operations, the fundamental tension between operational efficiency and democratic accountability has become acute. [1] This survey article provides a comprehensive overview of the evolving landscape of Explainable Artificial Intelligence (XAI) in Decision Support Systems, examining how transparency, interpretability, and trust become paramount as AI plays an increasingly crucial role in decision-making processes. [2] Transparency in AI decision-making has emerged as a pivotal research area, striving to augment transparency and interpretability of AI systems in critical domains such as healthcare, finance, criminal justice, and beyond.
The core challenge underpinning government AI deployment is what might be termed the &quot;explainability-performance paradox.&quot; [3] High-stakes decision-making in sectors like healthcare, finance, and criminal justice requires both exceptional accuracy and clear interpretability. However, the most performant models—deep learning architectures, ensemble methods, and complex feature interactions—are inherently opaque. [4] DARPA&#39;s seminal explainable AI program (2016-2021) identified this critical gap, establishing that while early AI systems employed symbolic reasoning that was inherently explainable, advances in machine learning created powerful but opaque models that cannot explain their decisions to human users, especially in mission-critical contexts where DoD and allied governments require comprehensible reasoning.
2. Explainable AI Methodologies and Technical Frameworks
2.1 Core XAI Techniques and Interpretability Methods
The XAI landscape encompasses diverse technical approaches that can be categorized along several dimensions. [5] A theoretical framework for AI interpretability categorizes explanation methods based on scope (local vs. global), timing (post-hoc vs. intrinsic), and dependency on model architecture. Key techniques include LIME (Local Interpretable Model-agnostic Explanations) and SHAP (SHapley Additive exPlanations), which provide local and global interpretability respectively. [6] A hybrid XAI framework integrating rule-based models with deep learning employs Layer-Wise Relevance Propagation (LRP) and SHAP for interpretability, achieving 94.3% accuracy while maintaining explainability and reaching a trustworthiness index of 92.1%.
The distinction between post-hoc and intrinsically interpretable models remains foundational. [7] XAI aims to provide a suite of machine learning techniques that enable human users to understand, appropriately trust, and produce explainable results. The taxonomy distinguishes between approaches targeting different phases of development: generating explanations, psychological models of explanation, and evaluation of explanation quality. [8] This comprehensive study examines explainability through four axes: data explainability, model explainability, prediction explainability, and explanation evaluation, across 410 peer-reviewed articles published between January 2016 and October 2022.
2.2 Domain-Specific Application and Calibration
XAI effectiveness varies substantially by application domain, requiring domain-specific calibration. [9] Recent AI-based methodological development shows that while large numbers of ML and DL models are inherently complex and lack explanations of their decision-making process, termed &quot;Black-Box,&quot; this opacity remains a major bottleneck for adoption in mission-critical domains. [10] In healthcare contexts specifically, XAI bridges the gap between model performance and human interpretability through post-hoc methods like SHAP and LIME, as well as intrinsically interpretable models, with persistent challenges including balancing explanation fidelity with usability and addressing data biases.
[11] A systematic review analyzing 68 articles from healthcare reveals that while most existing studies employ post-hoc, model-agnostic approaches such as SHAP and Grad-CAM, clinician sample sizes remain below 25 participants, and explanations generally improve trust and diagnostic confidence but frequently increase cognitive load and exhibit misalignment with clinical reasoning. [12] The clinical adoption challenge stems from the gap between technical XAI solutions and real-world clinician needs, requiring user-centered approaches that support workflow integration, calibrated trust, and robust evaluation methodologies capturing clinician-AI interaction patterns.
3. AI Governance, Accountability, and Regulatory Frameworks
3.1 Strategic AI Governance Architecture
The integration of AI into government operations demands comprehensive governance frameworks that extend beyond technical compliance. [13] A practical, risk-based governance model for public sector AI classifies systems into low-, medium-, and high-risk uses, aligning safeguards to impact. For high-risk systems such as eligibility decisions, law-enforcement support, and biometric identification, minimum requirements include named accountability ownership, meaningful human oversight, pre-deployment impact assessment, proportionate explainability, data quality and fairness testing, security controls with audit trails, enforceable procurement clauses for vendor accountability, and accessible grievance and review mechanisms.
[14] AI governance should be treated as a strategic governance function anchored in board oversight and enterprise risk management. The AI Governance Strategic Framework (AIGSF) connects ethical accountability, regulatory readiness, cybersecurity resilience, and performance outcomes, with case illustrations across hiring, credit, consumer services, and generative AI demonstrating controls such as model documentation, algorithmic audits, impact assessments, and human-in-the-loop oversight.
[15] XAI has emerged as a critical scientific advancement to reconcile predictive performance with regulatory oversight requirements in automated compliance and policy enforcement. Contemporary XAI models designed for regulatory compliance include intrinsically interpretable models (decision trees, generalized additive models, attention-based architectures), post-hoc explanation tools (SHAP, LIME, counterfactual reasoning), and hybrid architectures integrating symbolic rule engines with deep learning classifiers to align automated outputs with statutory language.
3.2 Cross-Sector Governance and Institutional Harmonization
[16] Secure AI governance in healthcare requires compliance with global regulations such as HIPAA, GDPR, and FDA AI/ML guidelines, integrating data governance, model lifecycle management, and auditability mechanisms to ensure end-to-end trust and transparency. Key components include data lineage tracking, bias mitigation, automated compliance monitoring, and federated learning for privacy preservation. [17] Industrial governance frameworks must embrace a five-dimensional trust model spanning quality, security, privacy, fairness, and explainability across data, service, and knowledge governance layers, addressing semantic interoperability, runtime policy enforcement, and operational-IT alignment.
[18] Organizations combining explainable AI with empowered ethics boards experience significantly fewer governance failures, with 48% fewer instances of bias and regulatory violations, and 35% fewer regulatory investigations when third-party audits are employed. However, ethics boards with only advisory roles show limited impact, highlighting the necessity for real effectiveness to stem from integration into core decision-making processes.
[19] A five-layer AI governance framework spans from broad regulatory mandates to specific standards, assessment methodologies, and certification processes, providing a structured pathway to meet technical, regulatory, and ethical requirements. By narrowing scope through progressively focused layers, the framework addresses gaps in legal mandates, standardization, and implementation, demonstrating applicability to both global and region-specific AI governance needs.
4. Provenance, Traceability, and Data Governance Systems
4.1 Data Provenance and Lineage Architecture
Establishing trustworthy AI systems requires comprehensive, immutable documentation of data lineage and model provenance. [20] Engineering trust in AI systems through data-layer controls creates explainability and auditability via provenance capture, lineage graphs, feature-store governance, and standardized documentation artifacts. The framework produces machine-checkable evidence of lifecycle activities from ingestion to inference, supporting inspection and defensible documentation of model behavior in production.
[21] AI Commit Ledger, a local-first, Git-native provenance system, generates structured AI Receipts for each commit, achieving 94.2% attribution accuracy with less than 340ms overhead per commit, enabling organizations to enforce AI-related policies, support post-incident analysis, and satisfy regulated reporting requirements. [22] Blockchain-powered data provenance provides a comprehensive exploration of integrating distributed ledger systems with cryptographic verification, zero-knowledge proofs, and federated logging to ensure verifiability without exposing sensitive data, with simulation experiments demonstrating that blockchain-enabled audits improve transparency and reduce fraudulent activities by more than 50% compared to traditional audit approaches.
4.2 Transparency Mechanisms and Audit Trails
[23] The AI Product Passport provides a standards-based framework improving transparency, traceability, and compliance through lifecycle-based documentation, with automated provenance tracking and role-based access. Implementation generates machine- and human-readable reports customizable for diverse stakeholders, aligning with FUTURE-AI principles (Fairness, Universality, Traceability, Usability, Robustness, Explainability). [24] An auditable, source-verified framework for clinical AI decision support integrates curated medical knowledge bases with explicit provenance metadata and retrieval-augmented generation engines that link recommendations to identifiable guidelines and peer-reviewed sources, with tamper-evident audit logging capturing system inputs, retrieved evidence, and inference steps.
[25] XAI-Compliance-by-Design introduces a modular engineering framework routing explainability outputs into structured, audit-ready evidence aligned with EU AI Act and GDPR obligations. The framework specifies modular architecture separating technical evidence generation from governance consumption through explicit interface points for emitting, storing, and querying evidence, alongside a Technical-Regulatory Correspondence Matrix mapping regulatory anchors to concrete evidence artifacts.
[26] Provenance-based auditing successfully detects gender biases in clinical models, with logistic regression exhibiting statistically significant bias (EOD = +0.256, p = 0.0080) while random forest&#39;s smaller disparity was not statistically significant, demonstrating how structured provenance records enable fairness audits and support regulatory compliance. The AI Fairness Provenance Record documents data origin, model choices, and bias metrics, enabling auditors to trace decisions to their source.
5. Assurance Cases and Certification Frameworks for AI Systems
5.1 Assurance Case Development and Safety Arguments
Assurance cases provide structured arguments supported by evidence to justify key properties of systems, becoming increasingly essential for autonomous and AI-enabled systems. [27] An Overarching Properties-based approach for aerospace systems uses structured premise-based arguments addressing AI/ML technologies with respect to system-level safety and design needs. The Overarching Properties labeled Intent, Correctness, and Innocuity align low-level properties of AI/ML components to system-level properties, making it easy to logically establish safety and correctness of AI-based digital aerospace systems.
[28] The Landscape of AI Safety Concerns methodology systematically supports creation of safety assurance cases by demonstrating absence of AI safety concerns. The methodology addresses the challenge of assuring safety in systems incorporating AI components, where lack of concrete specifications and system complexity lead to uncertain behavior, requiring systematic analysis and mitigation of AI-specific insufficiencies to yield essential evidence supporting convincing assurance cases.
[29] Continuous Assurance Framework integrates design-time, runtime, and evolution-time assurance within a traceable, model-driven workflow using formal verification methods (RoboChart for functional correctness, PRISM for probabilistic risk analysis). The approach automatically regenerates structured assurance arguments whenever formal specifications or verification results change, ensuring traceability and addressing the challenge of sustaining justified confidence in autonomous systems&#39; correctness across their operational lifecycle.
5.2 Ethics Assurance and Principles-Based Arguments
[30] A principles-based ethics assurance argument pattern (PRAISE) brings together assurance case methodology with ethical principles (justice, beneficence, non-maleficence, respect for human autonomy, with transparency as supporting role) to structure ethics assurance cases. PRAISE provides a reusable template enabling engineers, developers, operators, or regulators to justify, communicate, or challenge claims about overall ethical acceptability of specific AI/AS use in given socio-technical contexts.
[31] The dependability perspective on AI assurance outlines principles of classical assurance for computer-based systems, examining application to systems employing AI and ML. This perspective aims to minimize trust in AI/ML elements by using &quot;defense in depth&quot; with a hierarchy of less complex systems to &quot;guard&quot; AI components, which contrasts with the &quot;trustworthy&quot; perspective seeking to apply assurance to AI/ML elements themselves. For guards requiring perception using AI/ML, methods to minimize trust include diversity, defense in depth, explanations, and micro-Operational Design Domains (ODDs).
[32] Trustworthiness Assurance Assessment for high-risk AI systems proposes methodologies achieving compliance with the EU AI Act through seven interconnected requirements achieved throughout the AI lifecycle. The operational design domain and behavior competency concepts from autonomous driving are utilized in risk assessment strategies to quantify different types of residual risks, with sufficient consideration of human factors and robust implementation across the AI lifecycle.
6. Government and Defense-Specific Applications
6.1 Public Sector AI Deployment and Democratic Governance
[33] AI governance in public sector enterprise systems requires foundational frameworks managing challenges related to transparency, accountability, legal compliance, and ethical responsibility. Public sector organizations operate under heightened scrutiny where automated decisions must be explainable, fair, and aligned with democratic values and regulatory obligations. Governance mechanisms spanning strategic policy alignment, operational oversight, and technical controls ensure trustworthy and compliant AI deployment across system lifecycles.
[34] A conceptual model for responsible AI in public-facing digital services comprises five interlocking components: stakeholder-centered transparency, regulatory compliance, human oversight, adaptive feedback loops, and equity preservation. This model ensures that trustworthy practices integrate into platform mechanisms, supporting trust, user agency, and equity, with inclusive stakeholder engagement at every stage alongside policy instruments supporting continuous evaluation and redress mechanisms.
[35] Predictive analytics frameworks in public project management enable real-time decision support, with practical implementations demonstrating 18% reduction in eldercare backlog waitlists within nine months, reduction in crime response lead times from 14 days to under three hours, and improvement in budget variance accuracy from ±15% to ±4%. Such frameworks support data-driven decisions while remaining compliant with public sector regulations, privacy constraints, and equity goals.
6.2 Defense and Military AI Governance
[36] Military decision contexts present unique trust challenges, requiring calibrated trust via training, transparency, human-machine teaming, and robust data governance. Short-term hazards include skill fade and corrupted data, while long-term risks involve adversarial tampering. Policy and educational interventions must preserve critical thinking, assure data integrity, and maintain human agency in AI-augmented warfare through sustained collaboration between technical specialists and operational leaders.
[37] AI-driven risk management for critical infrastructure and national security projects shifts from reactive to proactive risk management through predictive analytics, autonomous risk mitigation, and real-time decision support. Case studies across cybersecurity, supply chain management, and national security operations demonstrate AI&#39;s effectiveness in reducing vulnerabilities and optimizing risk mitigation efforts, though adoption must be guided by robust governance policies, ethical standards, and regulatory compliance measures.
[38] Military health applications of AI range from AI-driven telementoring and decision support tools to automated trauma documentation and remote patient monitoring, with specific initiatives enhancing military trauma care through explainable AI systems. Implementation faces obstacles including robust data collection methods, secure and interoperable storage solutions, and frameworks addressing ethical and trust issues, with decentralized storage technologies like blockchain and explainable AI proposed to enhance reliability and transparency.
7. Critical Challenges, Gaps, and Future Directions
7.1 Persistent Implementation and Adoption Barriers
Despite substantial advances in XAI methodology and governance frameworks, significant barriers persist. [39] Over 80% of clinical studies adopt post-hoc, model-agnostic approaches with clinician sample sizes below 25 participants, indicating explanations generally improve clinician trust but frequently increase cognitive load and misalign with domain reasoning. [40] Empirical analysis of XAI techniques reveals that while XAI methods improve interpretability and stakeholder trust, they often come with computational and accuracy trade-offs, with the study identifying challenges and opportunities in integrating XAI into real-world applications.
[41] Gaps in AI-compliant governance frameworks emerge particularly for low-capacity actors—especially SMEs and public authorities—with persistent issues of enforceability, proportionality, and auditability compounded by frictions between regulatory instruments like the EU AI Act and GDPR. Compliance asymmetry, the structural friction between regulatory ambition and institutional capacity, characterizes many jurisdictions&#39; implementation challenges.
7.2 Regulatory and Standardization Fragmentation
[42] ISO standards such as ISO/IEC 24027 and 24368 aim to embed fairness, explainability, and risk control into AI development, though effectiveness varies across legal and policy landscapes. Alignment with EU risk-tiered AI Act, China&#39;s social stability focus, and the U.S. decentralized regulatory model reveals that compliance often lacks enforceability and fails to accommodate local values, with recommendations for mandatory audits, region-specific annexes to standards, and integrated privacy-risk modules.
[43] The roadmap to developing legal policies for AI deployment reveals that stringent regulations enhance clarity and foster public trust but impose higher compliance costs and hinder innovation, while lenient approaches promote innovation but create legal ambiguity. Key challenges including standards harmonization, algorithm transparency, and accountability remain critical issues requiring international collaboration to establish robust, secure, and sustainable governance frameworks.
[44] Comparative analysis of AI frameworks in the US, EU, Singapore, and UAE reveals substantial differences in AI governance approaches, compliance challenges, and privacy protection gaps across regions, highlighting the importance of harmonized standards and proposing mutual recognition agreements (MRAs) and alignment with OECD AI Principles to promote uniform global standards.
7.3 Emerging Research Gaps and Methodological Challenges
[45] Human-centric AI governance requires an adaptive public international law framework addressing ethical accountability, regulatory adaptability, and transparency. Existing models like WHO&#39;s lack enforceability while GDPR&#39;s rigid data-sharing rules hinder interoperability, necessitating scalable, region-specific regulations enhancing interoperability while respecting local governance needs through modular regulation and accountability mechanisms.
[46] AI governance in cybersecurity and IT risk domains must establish strong governance frameworks balancing innovation with accountability through principles of transparency, robustness, and regulatory compliance. Findings suggest that governance frameworks are important for maintaining trust, mitigating operational risks, and achieving sustainable, ethical AI deployment in critical IT infrastructures through systematic threat metrics, adversarial hardening, and real-time anomaly detection.
[47] Putting AI ethics to work reveals gaps in auditing practices that should be considered going forward, with assessment of current frameworks showing that while best practices from historical risk assessments in other domains provide comparison typologies, current AI ethics evaluations often fall short in supporting diverse stakeholders across development lifecycles and types of audits.
Conclusion
The deployment of explainable AI in government decision-support systems represents a fundamental shift in how public institutions can balance technological advancement with democratic accountability. The convergence of XAI methodologies, robust governance frameworks, comprehensive provenance systems, and formal assurance cases provides governments with essential tools to maintain oversight and transparency. Yet significant implementation challenges persist, particularly regarding standardization across jurisdictions, capacity constraints in resource-limited administrations, and the persistent tension between model performance and interpretability.
Successful government AI deployment requires integration across technical, governance, and institutional dimensions. Organizations must embed explainability at the core of AI lifecycle management, not as an afterthought, while establishing governance structures that maintain meaningful human oversight and create accessible redress mechanisms for affected citizens. The evidence consistently demonstrates that transparent, auditable, and ethically grounded AI systems achieve superior outcomes in terms of stakeholder trust, regulatory compliance, and operational effectiveness.
As government AI adoption accelerates, particularly in defense and critical infrastructure contexts, the frameworks, methodologies, and governance approaches discussed in this review provide essential guidance for responsible deployment. However, achieving widespread, trustworthy government AI requires sustained commitment to international harmonization of standards, investment in institutional capacity building, particularly in developing economies, and ongoing research addressing the fundamental challenges of maintaining human agency and democratic values in increasingly AI-augmented governance environments.
References
[1] 	G. Kostopoulos, G. Davrazos, and S. Kotsiantis, “Explainable artificial intelligence-based decision support systems: A recent review,” Electronics, July 2024, doi: 10.3390/electronics13142842.
[2] 	J. R, “Transparency in AI decision making: A survey of explainable AI methods and applications,” Advances in Robotic Technology, Jan. 2024, doi: 10.23880/art-16000110.
[3] 	R. David, H. Shankar, P. Kura, K. Kowtarapu, U. S, and S. Karkuzhali, “Advancement in explainable AI: Bringing transparency and interpretability to machine learning models for use in high-stakes decisions,” 2025 International Conference on Emerging Smart Computing and Informatics (ESCI), Mar. 2025, doi: 10.1109/ESCI63694.2025.10988079.
[4] 	D. Gunning, E. S. Vorm, Y. Wang, and M. Turek, “<Scp>DARPA</scp>”s explainable<scp>AI</scp>(<scp>XAI</scp>) program: A retrospective,” Applied AI Letters, Dec. 2021, doi: https://doi.org/10.1002/ail2.61.
[5] 	A. Chinnaraju, “Explainable AI (XAI) for trustworthy and transparent decision-making: A theoretical framework for AI interpretability,” World Journal of Advanced Engineering Technology and Sciences, Mar. 2025, doi: 10.30574/wjaets.2025.14.3.0106.
[6] 	B. K. Malamuthu, T. P. Balakrishnan, J. Deepika, N. P, B. Venkataramanaiah, and V. Malathy, “Explainable AI for decision-making: A hybrid approach to trustworthy computing,” International Journal of Computational and Experimental Science and Engineering, May 2025, doi: 10.22399/ijcesen.1684.
[7] 	R. Dwivedi et al., “Explainable AI (XAI): Core ideas, techniques, and solutions,” ACM Computing Surveys, Sept. 2022, doi: https://doi.org/10.1145/3561048.
[8] 	S. Ali et al., “Explainable artificial intelligence (XAI): What we know and what is left to attain trustworthy artificial intelligence,” Information Fusion, Apr. 2023, doi: https://doi.org/10.1016/j.inffus.2023.101805.
[9] 	V. Hassija et al., “Interpreting black-box models: A review on explainable artificial intelligence,” Cognitive Computation, Aug. 2023, doi: https://doi.org/10.1007/s12559-023-10179-8.
[10] 	O. Ozdemir, “Explainable AI (XAI) in healthcare: Bridging the gap between accuracy and interpretability,” Journal of Science, Technology and Engineering Research, June 2024, doi: 10.64206/0z78ev10.
[11] 	N. A. Aziz, A. Manzoor, M. Deedahwar, M. Qureshi, and W. Rashwan, “Explainable AI in healthcare: Systematic review of clinical decision support systems,” medRxiv, Aug. 2024, doi: 10.1101/2024.08.10.24311735.
[12] 	M. Salimparsa, K. Sedig, D. Lizotte, S. S. Abdullah, N. Chalabianloo, and F. Muanda, “Explainable AI for clinical decision support systems: Literature review, key gaps, and research synthesis,” Informatics, Oct. 2025, doi: 10.3390/informatics12040119.
[13] 	N. Ahmad, “AI-ENABLED PUBLIC GOVERNANCE IN DEVELOPING STATES: SERVICE DELIVERY GAINS, ACCOUNTABILITY RISKS, AND a PRACTICAL RISK-BASED REGULATORY MODEL,” Lex localis - Journal of Local Self-Government, Jan. 2026, doi: 10.52152/wja5db40.
[14] 	D. Martins, “Artificial intelligence governance in corporate strategy: Ethical risk, regulatory compliance, and competitive advantage,” International Journal of Business &amp; Management Studies, Mar. 2026, doi: 10.56734/ijbms.v7n3a4.
[15] 	C. Aliliele, I. S. Mbonu, E. Uzoka, and U. Iwuanyanwu, “Explainable artificial intelligence models for automated regulatory compliance and policy enforcement,” Shodhshauryam International Scientific Refereed Research, June 2025, doi: 10.32628/shisrrj258474.
[16] 	S. Lanka, “Secure AI governance in healthcare: Ensuring compliance, auditability, and data trust across the ML lifecycle,” International journal of science research and technology, Mar. 2026, doi: 10.15662/ijsrat.2023.0605006.
[17] 	H. Zhao et al., “Industrial data-service-knowledge governance: Toward integrated and trusted intelligence for industry 5.0,” arXiv.org, Jan. 2026, doi: 10.48550/arXiv.2601.04569.
[18] 	R. Nakajima, “Bridging the AI governance gap: Evaluating the effectiveness of transparency tools and ethics boards in multinational firms,” Corporate governance: International outlook, 2025, doi: 10.22495/cgiop4.
[19] 	A. Agarwal and M. Nene, “A five-layer framework for AI governance: Integrating regulation, standards, and certification,” Transforming Government: People, Process and Policy, May 2025, doi: 10.1108/TG-03-2025-0065.
[20] 	M. A. Shareef, “Engineering trust in AI systems: A data-layer framework for explainability and auditability,” The American Journal of Interdisciplinary Innovations and Research, Feb. 2026, doi: 10.37547/tajiir/volume08issue02-11.
[21] 	P. Gaikwad, “AI commit ledger: Bridging the AI transparency gap in version control with deterministic session attribution,” International Journal of Advanced Research in Science, Communication and Technology, Mar. 2026, doi: 10.48175/ijarsct-31627.
[22] 	Prof. (Dr. ). A. Jain, “Blockchain-powered data provenance for AI model audits,” Scientific Journal of Artificial Intelligence and Blockchain Technologies, Jan. 2024, doi: 10.63345/sjaibt.v1.i1.104.
[23] 	A. A. Sinaci et al., “Enhancing transparency and traceability in healthcare AI: The AI product passport,” arXiv.org, Dec. 2025, doi: 10.48550/arXiv.2512.13702.
[24] 	F. F. Alu and S. Oluwadare, “An auditable and source-verified framework for clinical AI decision support: Integrating retrieval-augmented generation with data provenance,” Frontiers in Artificial Intelligence, Feb. 2026, doi: https://doi.org/10.3389/frai.2026.1737532.
[25] 	A. Gonçalves and A. Correia, “Engineering explainable AI systems for GDPR-aligned decision transparency: A modular framework for continuous compliance,” Journal of Cybersecurity and Privacy, Dec. 2025, doi: 10.3390/jcp6010007.
[26] 	F. Alu, S. Oluwadare, N. Halliday, and U. E. Agwunobi, “Auditing fairness in clinical AI systems using provenance-based simulation: A comparative and regulatory perspective,” Frontiers in Artificial Intelligence, Apr. 2026, doi: 10.3389/frai.2026.1756023.
[27] 	S. Paul et al., “Assurance of AI/ML-based aerospace systems using overarching properties,” Symposium on Dependable Autonomic and Secure Computing, Sept. 2024, doi: 10.1109/DASC62030.2024.10749102.
[28] 	R. Schnitzer, L. Kilian, S. Roessner, K. Theodorou, and S. Zillner, “Landscape of AI safety concerns - a methodology to support safety assurance for AI-based autonomous systems,” International Conference on System Reliability and Safety, Nov. 2024, doi: 10.1109/ICSRS63046.2024.10927556.
[29] 	D. B. Abeywickrama, M. Fisher, F. Wheeler, and L. A. Dennis, “Towards continuous assurance with formal verification and assurance cases,” arXiv.org, Nov. 2025, doi: 10.48550/arXiv.2511.14805.
[30] 	Z. Porter, I. Habli, J. Mcdermid, and M. H. L. Kaas, “A principles-based ethics assurance argument pattern for AI and autonomous systems,” AI and Ethics, Mar. 2022, doi: 10.1007/s43681-023-00297-2.
[31] 	R. Bloomfield and J. Rushby, “Assurance of AI systems from a dependability perspective,” arXiv.org, July 2024, doi: 10.48550/arXiv.2407.13948.
[32] 	G. Stettinger, P. Weissensteiner, and S. Khastgir, “Trustworthiness assurance assessment for high-risk AI-based systems,” IEEE Access, 2024, doi: 10.1109/ACCESS.2024.3364387.
[33] 	J. Bhat, D. Sundar, and Y. Jayaram, “AI governance in public sector enterprise systems: Ensuring trust, compliance, and ethics,” International Journal of Emerging Trends in Computer Science and Information Technology, 2024, doi: 10.63282/3050-9246.ijetcsit-v5i1p114.
[34] 	B. A. Adewusi, B. I. Adekunle, S. D. Mustapha, and A. C. Uzoka, “A conceptual model for responsible AI integration in public-facing digital services and platform governance,” International Journal of Scientific Research in Computer Science Engineering and Information Technology, May 2024, doi: https://doi.org/10.32628/cseit2425416.
[35] 	O. C. Owens and P. Adeel, “Predictive analytics in public sector project management: A real-time decision support framework,” International journal of management and technology, Apr. 2025, doi: 10.37745/ijmt.2013/vol12n44457.
[36] 	F. Tudorache, “INCREASING TRUST IN ARTIFICIAL INTELLIGENCE IN MILITARY DECISION CONTEXTS. GENERATIONAL AND COGNITIVE IMPLICATIONS,” Studia Securitatis, Dec. 2025, doi: 10.54989/stusec.2025.19.02.13.
[37] 	Z. S. Munmun, M. D. Hossain, I. jahan, S. Akther, and A. A. Masum, “AI-driven project risk management: Leveraging artificial intelligence to predict, mitigate, and manage project risks in critical infrastructure and national security projects,” Journal of Computer Science and Technology Studies, Apr. 2025, doi: 10.32996/jcsts.2025.7.2.6.
[38] 	Z. Riggenbach, J. V. Eaton, A. Kelly, H. E. Kim, and J. R. Bingham, “AI on the front lines: A primer for the military health professional.” Military Medicine, May 2025, doi: 10.1093/milmed/usaf165.
[39] 	A. Gambetti, Q. Han, H. Shen, and C. Soares, “A survey on human-centered evaluation of explainable AI methods in clinical decision support systems,” arXiv.org, Feb. 2025, doi: 10.48550/arXiv.2502.09849.
[40] 	P. Verma, “Advancements in explainable AI: Bridging the gap between interpretability and performance in machine learning models,” International Journal of Machine Learning, AI &amp; Data Science Evolution, Jan. 2025, doi: 10.63665/ijmlaidse.v1i1.01.
[41] 	W. Finch and M. Butt, “Gaps in AI-compliant complementary governance frameworks” suitability (for low-capacity actors), and structural asymmetries (in the compliance ecosystem) - a systematic review,” Journal of Cybersecurity and Privacy, Nov. 2025, doi: 10.3390/jcp5040101.
[42] 	S. Sankaran, “Enhancing trust through standards: A comparative risk-impact framework for aligning ISO AI standards with global ethical and regulatory contexts,” None, Aug. 2025, doi: https://doi.org/10.1109/acdsa65407.2025.11166403.
[43] 	T. Trisnawati, “Artificial intelligence governance and regulation: A roadmap to developing legal policies for artificial intelligence deployment,” Journal of Governance and Administrative Reform, Dec. 2024, doi: https://doi.org/10.20473/jgar.v5i2.65194.
[44] 	A. N. F. Bahgat, “AI governance and data privacy in cross-border contexts: A comparative analysis of regulatory frameworks,” Journal of Data Protection &amp; Privacy, Sept. 2025, doi: 10.69554/bojl3033.
[45] 	F. S. Sedeeq and P. Arman, “Human-centric AI governance: An adaptive public international law framework for ethical and inclusive AI regulation in public health,” Journal of Law, Medicine & Ethics, Sept. 2025, doi: 10.1017/jme.2025.10175.
[46] 	A. Punia, “AI governance models for secure and ethical deployment in cyber security and IT risk environments,” 2025 IEEE 1st International Conference on Recent Trends in Computing and Smart Mobility (RCSM), Dec. 2025, doi: 10.1109/RCSM67767.2025.11506917.
[47] 	J. Ayling and A. Chapman, “Putting AI ethics to work: Are the tools fit for purpose?” AI and Ethics, Sept. 2021, doi: https://doi.org/10.1007/s43681-021-00084-x.