Synthesizing the Future of AI-Blockchain Integration: A Pathway for Adaptive, Ethical, and Efficiency

LATIN-AMERICAN JOURNAL OF COMPUTING (LAJC), Vol XIII, Issue 1, January - June 2026 Synthesizing the Future of AI-Blockchain Integration: A Pathway for Adaptive, Ethical, and Efficiency Godwin Mandinyenya North-West University School of Computer Science and Information Systems Vaal Campus Vanderbijlpark, South Africa [email protected] ORCID: 0009-0001-7659-4400 Vusimuzi Malele North-West University School of Computer Science and Information Systems Vaal Campus Vanderbijlpark, South Africa [email protected] ORCID: 0000-0001-6803-9030 Abstract— This study systematically examines the transformative role of Artificial Intelligence (AI) in addressing the persistent challenges of blockchain technology across protocols, smart contracts, and distributed ledger management. Although blockchain offers decentralization, immutability, and transparency, its broader adoption remains constrained by scalability limitations, security vulnerabilities, inefficient consensus mechanisms, and the complexity of contract design and auditing. The findings of this review demonstrate that AI provides promising solutions to these barriers. Reinforcement learning (RL) applied to Proof-of-Stake reduced consensus latency by 30-50%, while NLP-based smart contracts lowered vulnerabilities by up to 40%, though both approaches introduced new concerns related to energy overheads and auditability. In addition, intelligent algorithms enhance ledger efficiency and data analytics, supporting more scalable and secure transaction processing. This Drawing on 28 peerreviewed studies published between 2018 and 2024 and guided by the PRISMA 2020 framework, this paper synthesizes state-of-the-art research, maps sector-specific applications in finance, healthcare, and supply chain management, and highlights unresolved gaps in ethics, reproducibility, and regulatory compliance. Notably, only 12% of the reviewed studies validated their approaches on live networks underscoring the gap between simulationdriven research and real-world deployment. The discussion culminates in the AI–Blockchain Interaction Model (AIBIM), a conceptual framework that systematizes synergies across consensus, contract, and application layers. By integrating empirical insights with critical evaluation, this work emphasizes the interdisciplinary nature of AI– blockchain research and provides actionable directions for advancing decentralized, scalable, and ethically aligned systems. This synthesis provides actionable insights for developers, regulators, and researchers in deploying AIblockchain systems across finance, healthcare, and supply chains. Keywords— Blockchain, Artificial Intelligence, Smart Contracts, Consensus Mechanisms, Distributed Ledger, Deep Learning, Formal Verification I. INTRODUCTION Blockchain technology has emerged as a groundbreaking innovation capable of transforming diverse industries by providing decentralized, immutable, and transparent infrastructures for data storage and transaction processing [23]. Its applications span finance, healthcare, supply chain management, and governance, where distributed ledgers are increasingly viewed as enablers of trust and accountability [6], [16], [24]. However, the widespread adoption of blockchain remains constrained by persistent challenges, including scalability bottlenecks, security vulnerabilities, the inefficiency of consensus mechanisms, and the complexity of smart contract creation and auditing [13], [19]. Artificial Intelligence (AI) has been identified as a promising solution to many of these limitations [1], [4]. By leveraging machine learning and predictive analytics, AI can enhance blockchain protocols through the optimization of consensus algorithms, leading to faster transaction finalization and improved fault tolerance [5], [7]. AI-based anomaly detection techniques, such as graph neural networks (GNNs), further strengthen network resilience by identifying malicious activity, including 51% attacks, with high accuracy [3], [21]. In the realm of smart contracts, AI contributes to greater automation and reliability. Natural Language Processing (NLP) techniques have been used to generate and audit contracts directly from textual requirements, reducing vulnerabilities and improving execution accuracy [4], [22]. Supervised learning and explainable AI (XAI) methods also offer the potential to identify flaws in contract logic, thereby minimizing risks associated with opaque, non-interpretable models [12], [18]. AI can also improve the efficiency of distributed ledgers, where intelligent algorithms optimize storage, retrieval, and compression processes [11], [13]. Such approaches enable more scalable and sustainable blockchain systems by reducing storage overheads and facilitating advanced data analytics for informed decision-making [25], [26]. These innovations indicate that the synergy between AI and blockchain represents not just incremental improvement, but a paradigm shift toward robust, adaptive, and intelligent decentralized systems [7], [17]. G. Mandinyenya and Vusimuzi Malele, “Synthesizing the Future of AI-Blockchain Integration: A Pathway for Adaptive, Ethical, and Efficiency”, Latin-American Journal of Computing (LAJC), vol. 13, no. 1, 2026. This paper systematically examines how AI is being integrated into blockchain technologies to overcome fundamental limitations. Using the PRISMA 2020 framework, it reviews 28 peer-reviewed studies published between 2018 and 2024 to analyze contributions across protocols, smart contracts, and sector-specific applications. In doing so, the study also identifies critical gaps in reproducibility, ethical and legal integration, and sectoral diversity. To address these, the paper introduces the AI– Blockchain Interaction Model (AIBIM), a conceptual framework that systematizes synergies across consensus, contract, and application layers. By combining empirical evidence with conceptual innovation, this work provides actionable insights for developers, policymakers, and researchers seeking to advance the next generation of decentralized intelligence. A. Research Objectives • How can AI enhance blockchain protocols, smart contracts, and ledger efficiency? • What are the technical benefits and challenges of AIblockchain integration? • What sector-specific use cases demonstrate AIdriven blockchain optimisation? • What future advancements are anticipated in AIblockchain synergy? • What ethical and legal risks emerge from AIaugmented blockchain systems? • Propose a conceptual model to systematize interactions between AI and blockchain components. B. Contributions of the study This study provides a systematic analysis of the interdependencies between AI and blockchain technologies, highlighting how their integration reshapes protocols, smart contracts, and ledger management. The review identifies quantifiable improvements introduced by AI, including enhanced consensus performance, automated contract verification, and optimized storage techniques. In addition to these technical contributions, the findings showcase novel application domains across industries such as finance, healthcare, and supply chain management, underscoring the transformative potential of decentralized intelligence. At the same time, the review acknowledges several technical and implementation barriers, including energy tradeoffs in AI-enhanced consensus, the opacity of noninterpretable models in smart contracts, and the scalability limits of AI-based storage solutions. To address these challenges, the study outlines regulatory risks and corresponding mitigation strategies, such as the use of zeroknowledge proofs to support GDPR compliance and hybrid arbitration frameworks to clarify liability in automated contracts. Finally, the research contributes a validated conceptual model for AI–blockchain integration—the AI–Blockchain Interaction Model (AIBIM)—which systematizes synergies across consensus, contract, and application layers. This model not only synthesizes the evidence reviewed but also provides a structured roadmap for advancing secure, efficient, and ethically aligned AI–blockchain systems. II. LITERATURE REVIEW The fusion of artificial intelligence (AI) and blockchain technology is redefining decentralized systems by enhancing scalability, security, and automation [9]. This review critically examines advancements in AI-driven blockchain protocols, smart contracts, and sectoral implementations while highlighting unresolved ethical and technical challenges [28]. A. AI-Driven Blockchain Protocol Optimization AI enhances blockchain protocols by optimizing consensus mechanisms, security, and scalability. Reinforcement learning (RL) dynamically adjusts validator selection in Proof-of-Stake (PoS) systems, reducing consensus latency by 30-50%, though energy costs for AI training offset 20-25% of gains [1, 5] Graph Neural Networks (GNNs) detect malicious nodes and 51% attacks with >99% accuracy, while Federated Learning enables privacypreserving, decentralized AI training, reducing cross-shard communication by 35% in Hyperledger Fabric. However, 80% of studies test protocols on synthetic networks, neglecting real-world variables like node churn [3], [4]. However, most of these contributions are validated in simulated environments, limiting their external validity. The absence of large-scale, real-world pilots raises concerns about how well such optimizations would perform under heterogeneous network conditions or adversarial settings. Beyond protocols, AI also transforms smart contract development, where automation and explainability are central. B AI-Enhanced Smart Contracts AI automates smart contract development and auditing. Natural Language Processing (NLP) models generate Solidity code from plain text, reducing manual errors by 35%, but AI-generated code introduces novel vulnerabilities. Hybrid human-AI auditing tools achieve 95% accuracy in detecting re-entrancy bugs but miss 15% of logic flaws. Machine learning enables context-aware contracts (e.g, LSTM models adjusting DeFI interest rates), improving loan repayment rates by 20%. However, black-box AI models (e.g., deep neural networks) hinder auditability, raising compliance risks in regulated sectors. While these methods show high accuracy in controlled tests, their reliance on synthetic datasets and simulated blockchain testbeds means their reliability in production systems, such as Ethereum mainnet, remains uncertain. This limitation underscores the broader challenge of reproducibility in AI-blockchain research. C. Sector – Specific Implementations LATIN-AMERICAN JOURNAL OF COMPUTING (LAJC), Vol XIII, Issue 1, January - June 2026 . • Finance: AI predicts DeFI liquidity risks (25% lower impermanent loss) and optimises crossborder payments (settlements in minutes) [9]. • Healthcare: FL-trained models on blockchain achieve 98% diagnostic accuracy while complying with GDPR [6]. • Supply Chain: AI optimises IoT-blockchain logistics, improving on-time shipments by 30%. Agriculture and energy sectors remain underexplored, with only 3% of studies addressing these domains [12, 25]. By contrast, domains such as agriculture and energy remain largely at the proof-ofconcept stage, with few studies moving beyond theoretical models or pilot simulations. This imbalance reinforces the sectoral bias in the literature and limits insights into how AI-blockchain integration might address sustainability challenges or resource management in underrepresented industries. Notably, fewer than 5% of studies addressed agriculture or energy applications, reinforcing the dominance of finance and healthcare. D. Ethical and Legal Challenges Privacy vs. Immutability: GDPR’s “right to be forgotten” conflicts with blockchain permanence; zero-knowledge proofs(ZKPs) anonymize data without altering ledger history [8]. Centralisation Risks: AI-optimised PoS networks concentrate power <10% of nodes, undermining decentralisation. Liability Gaps: No legal frameworks exist for AIinduced contract failures (e.g., $50M DeFI hacks from oracle errors) [10]. III. RESEARCH METHODOLOGY This study follows Petersen et al’s SLR framework. A. Planning Phase Research Goal To synthesise how AI enhances blockchain protocols, smart contracts, and efficiency, while identifying technical, sectorial, ethical, and legal implications integration. B. Research Questions (RQs) Formulated using PICOC (Population, Intervention, Comparison, Outcomes, Context): Final Research Questions (RQs): 1. RQ1: How can AI enhance blockchain protocols, smart contracts, and ledger efficiency? 2. RQ2: What are the technical benefits and challenges of AIblockchain integration? 3. RQ3: What sector-specific use cases demonstrate AI-driven blockchain optimisation? 4. RQ4: What future advancements are anticipated in AIblockchain synergy? 5. RQ5: What ethical and legal risks emerge from AIaugmented blockchain systems? 6. RQ6: How can interactions between AI and blockchain components be systematized? C. Search Strategy • Databases: IEEE Xplore, ACM Digital Library, Scopus, Web of Service, SpringerLink. • Search String: Designed using Boolean operators and tested for recall / precision: (artificial intelligence” OR “machine learning” OR “deep learning” OR “neural network) AND (“blockchain protocol” OR smart contract OR “distributed ledger” OR “consensus algorithm”) AND (“optimization” OR “efficiency” OR “security” OR “scalability”) • Timeframe: 2018-2024 (to capture post-secondgeneration blockchain advancements). Table 1 presents the inclusion and exclusion criteria applied in this review, ensuring that only peer-reviewed studies published between 2018 and 2024 with direct relevance to AI-blockchain integration were retained. TABLE 1: INCLUSION AND EXCLUSION CRITERIA Category Criteria Rationale Study Type Include: Primary studies (experiments, case studies). Secondary studies (reviews) excluded unless proposing novel frameworks. Exclude: Opinion pieces, nonpeer-reviewed preprints. Ensure methodological rigor and empirical validation. Blockchain In Focus Include: Papers where blockchain is central (e.g., protocols, smart contracts). Exclude tangential blockchain mentions (e.g., cryptocurrency price prediction). Include: Blockchain security / confidentiality papers only if AIintegrated. Aligns with RQs on AIdriven enhancements. AI Integration Include: Concrete AI techniques (e.g.,ML for consensus, Exclude theoretical AI models without G. Mandinyenya and Vusimuzi Malele, “Synthesizing the Future of AI-Blockchain Integration: A Pathway for Adaptive, Ethical, and Efficiency”, Latin-American Journal of Computing (LAJC), vol. 13, no. 1, 2026. NLP for contracts). blockchain implementation Fig.1. presents the PRISMA 2020 flow diagram, which outlines the systematic process followed in this review. From an initial pool of 1452 records across multiple databases, 312 duplicates were removed, followed by the title and abstract screening, and subsequent full-text assessment for eligibility. The diagram highlights how these stages ultimately narrowed the corpus to the final set of studies analysed, ensuring methodological transparency and adherence to systematic review best practice. D. PRISMA Flow Diagram Fig. 1. PRISMA Flow Diagram (Source, Author). Table 2 presents the coding scheme used to structure data extraction and align the reviewed evidence with the study’s research questions. AI Techniques (e.g., reinforcement learning, GNNs) were mapped to RQ1 and RQ2, reflecting their role in optimization and security. Blockchain Components (consensus, smart contracts, storage) were linked to RQ1 and RQ3 to capture modularity and performance trade-offs, while Performance Metrics (latency, throughput, accuracy) also addressed RQ1 and RQ2. Sectoral applications such as healthcare, finance, and supply chain corresponded to RQ3, highlighting domain-specific adoption patterns. Finally, Ethical and Legal Risks (bias, GDPR compliance, liability) informed RQ5, grounding the analysis in normative considerations. This coding framework ensured consistent categorization and guided synthesis across the review.” TABLE 2: CODING SCHEME / MAPPING VARIABLES TO RQs Variable Description Linked RQ AI Technique Reinforcement learning, GNNs RQ1, RQ2. Blockchain Component Consensus, smart contracts, storage RQ1, RQ3 Performance Metrics Latency, throughput, accuracy RQ1, RQ2` Sectoral Application Healthcare, finance, supply chain RQ3 Ethical / Legal Risks Bias, GDPR compliance, liability RQ5 Table 3 illustrates how the extracted data were systematically linked to the research questions. AI techniques in protocols were examined through frequency analysis of reinforcement learning versus GNN adoption, directly addressing RQ1 and RQ2. Sectoral use cases such as finance and healthcare were analyzed via thematic mapping to inform RQ3, while ethical risks including GDPR compliance and liability were assessed through content analysis, contributing to RQ5. This structured mapping ensured that each dimension of the dataset was coherently aligned with the study’s objectives and analytic strategy. TABLE 3: LINKING DATA TO RQs Data Type Analysis Method RQ Addressed AI Techniques in Protocols Frequency analysis of RL vs. GNN adoption RQ1, RQ2 Sectoral Use Cases Thematic mapping (finance vs. healthcare). RQ3 Ethical Risks Content analysis of GDPR / liability mentions. RQ5 Fig.2 illustrates the temporal distribution of the 28 included studies, showing steady growth between 2018 and 2020, followed by a sharp increase from 2021 onwards. This surge reflects the accelerating scholarly interest in AIblockchain integration, particulalrly in consensus optimization and smart contract automation. LATIN-AMERICAN JOURNAL OF COMPUTING (LAJC), Vol XIII, Issue 1, January - June 2026 . Fig.2. The temporal distribution of the 28 included studies Table 4 summarizes the quality assessment outcomes across the reviewed studies. The clarity of objectives scored highest, with an average of 4.2, indicating that 85% of papers explicitly articulated AI-blockchain research goals. Empirical validity received a moderate score of 3.8, reflecting that while 70% of studies relied on simulations, only 20% engaged with realworld data. Reproducibility was the weakest dimension, with an average score of 2.5, as just 15% of studies provided opensource code or datasets. These results highlight both the strengths in conceptual framing and the pressing need for more transparent and empirically validated contributions in AI-blockchain research. TABLE 4: QUALITY ASSESSMENT RESULTS Criterion Avg.Score (1-5) Key Findings Clarity of Objectives 4.2 85% explicitly addressed AIblockchain goals. Empirical Validity 3.8 70% used simulations; 20% realworld data. Reproducibility 2.5 Only 15% provided open-source code. IV RESULTS This systematic literature review synthesizes evidence from 28 peer-reviewed studies published between 2018 and 2024, with the aim of critically examining the transformative role of artificial intelligence (AI) in blockchain protocols, smart contracts, and sector-specific applications. Guided by the PRISMA 2020 framework and a mixed-methods analytical approach, the results are presented across three main dimensions. First, the review highlights technical innovations in AIdriven blockchain mechanisms, including reinforcement learning applied to consensus optimization [1], [5], graph neural networks (GNNs) for anomaly detection [3], and natural language processing (NLP) techniques for automated smart contract generation [4], [22]. These studies consistently demonstrate efficiency gains but also reveal new sources of vulnerability and resource overhead [7]. Second, sectorial applications are examined across finance, healthcare, and supply chain management. In finance, AI-enhanced DeFi systems improved liquidity risk prediction and transaction efficiency [24]. In healthcare, federated learning (FL) embedded in blockchain achieved diagnostic accuracy rates above 95% while ensuring GDPR compliance [6], [15]. Supply chain studies reported efficiency improvements of up to 30% in logistics optimization [16], though agriculture and energy remain underexplored [25], [26]. Despite promising results, most contributions rely on simulations rather than live deployments, which limits realworld generalizability. Third the analysis explores ethical and legal risks, particularly the tension between blockchain immutability and data privacy regulations such as the General Data Protection Regulation (GDPR) [8]. Other concerns include centralization tendencies in AI-controlled consensus [9], liability gaps in automated contracts [10], and the absence of robust regulatory frameworks [27], [28]. Collectively, these findings inform the development of the AI-Blockchain Interaction Model (AIBIM), a conceptual framework that systematizes AI-blockchain synergies across data, consensus, contract, and application layers. By integrating empirical evidence with critical evaluation, this framework provides actionable insights for developers, policymakers, and researchers seeking to advance secure, efficient, and ethically responsible decentralized systems. Table 5 categorizes the 28 studies according to their primary focus: protocol optimization, smart contracts, sectorspecific applications, and ethical/legal dimensions. The majority of contributions (22/28) emphasize protocol optimization, particularly reinforcement learning for consensus [1], [13], whereas ethical and legal considerations remain significantly underrepresented [27], [28]. TABLE 5: CATEGORISATION OF INCLUDED STUDIES (n=28) Cluster Count Key Focus Example Studies Performance Metrics Protocol optimisation 22 AIenhanced consensus, sharding, security [1] RL for PoS latency reduction 30–50% faster consensus; 25% lower energy use Smart Contracts 18 AIgenerated code, vulnerability detection, dynamic execution NLP for Solidity Code generation 40% fewer bugs; 20% faster deployment Sectoral Use Cases 15 Finance (DeFi), healthcare (data sharing), supply chain (IoT integration) Federated learning in healthcare blockchains 95% data accuracy; 60% storage reduction. G. Mandinyenya and Vusimuzi Malele, “Synthesizing the Future of AI-Blockchain Integration: A Pathway for Adaptive, Ethical, and Efficiency”, Latin-American Journal of Computing (LAJC), vol. 13, no. 1, 2026. Ethics / Legal 7 Bias in DAOs, GDPR conflicts, liability in AI-driven contracts [4] GDPRcompliance in immutable ledgers N/A (theoretical frameworks) The review revealed that protocol optimization dominated the literature, with 70% of studies (15 out of 22) focusing on enhancing consensus mechanisms such as Proof-of-Stake (PoS) and Practical Byzantine Fault Tolerance (PBFT). Reinforcement learning (RL) was the most widely applied approach, achieving latency reductions of 30–50% in 12 studies [1], [5], [13]. However, these improvements were often accompanied by increased energy demands, with some studies reporting up to 25% overhead during RL training [7]. In the area of smart contracts, supervised learning techniques were the most prevalent, appearing in 12 of the 18 studies reviewed [4], [14], [22]. These models demonstrated strong performance in vulnerability detection and automated contract generation, with detection accuracy exceeding 90%. Nevertheless, only three studies validated their methods on live blockchain networks such as Ethereum mainnet, underscoring a gap between experimental prototypes and production-grade applications. With respect to sectoral use cases, finance emerged as the leading application domain, accounting for two-thirds of the 15 studies identified [24]. Healthcare also featured prominently, particularly through federated learning for privacy-preserving diagnostics [6], [15]. By contrast, supply chain implementations were limited to only two studies [16], both of which lacked large-scale real-world validation. Other critical sectors such as energy and agriculture remained underexplored, represented in only isolated contributions [25], [26]. Finally, the ethical and legal dimension was the least developed, with all seven identified studies remaining at a theoretical level [8]– [10], [27], [28]. None provided actionable frameworks or empirical evaluations for addressing pressing concerns such as GDPR compliance, liability allocation, or bias in decentralized autonomous organizations (DAOs). Table 4 categorizes sectorial use cases, and Fig.3 further illustrates the distribution across industries, showing s strong dominance of finance and healthcare, while agriculture, energy, and governance remain marginally represented. This imbalance highlights the sectorial bias in current AIblockchain research and the need for broader application domains. Fig.3 Sectorial adoption of AI-Blockchain integration across 28 studies. Table 5 synthesizes the technical benefits and challenges of AI-blockchain integration. While AI-enhanced consensus mechanisms were shown to improve finalization speed by up to 60% [5], [21], they also introduced significant energy costs [7]. Similarly, AI-driven smart contracts enhanced bug detection accuracy [14], [22] but raised concerns around transparency and auditability, particularly when employing opaque deep learning models [12], [18]. As Fig. 3 shows, while latency reduction is significant, the trade-off is an unsustainable energy overhead. TABLE 5: TECHNICAL BENEFITS AND CHALLENGES OF AIBLOCKCHAIN INTEGRATION Componen t Benefits Challenge s Supportin g Studies Conflicting Evidence Consensus 40-60% faster finalisation (AI-PoS) High AI training overhead (25% energy cost) [5], [6] [7] reports 15% latency trade-off Smart Contracts 95% vulnerabilit y detection accuracy Black-box models reduce auditabilit y [8], [9] [10] finds 20% false positives Ledger Storage 60% compressio n via auto encoders Increased query latency (15–20%) [11], [12] [13] shows 30% compressio n loss over time Table 5 synthesizes the benefits and challenges of AIblockchain integration, particularly the trade-offs between efficiency and sustainability. Fig. 4 illustrates these tradeoffs, showing that while RL-optimized Proof-of-Stake reduces latency by up to 45%, it incurs an energy overhead of approximately 25%. In contrast, PBFT achieves moderate latency gains (30%) with a lower energy cost (15%). These results underscore the recurring tension between performance improvements and resource efficiency. LATIN-AMERICAN JOURNAL OF COMPUTING (LAJC), Vol XIII, Issue 1, January - June 2026 . Fig.4 Consensus performance gains versus energy overheads. The findings indicate that AI significantly enhances consensus mechanisms, particularly improving transaction speed and reducing latency. Reinforcement learning (RL) applied to Proof-of-Stake systems consistently improved consensus efficiency; however, these benefits were offset by resource costs, with RL training negating up to 25% of the performance gains [1], [5], [7]. This highlights the trade-off between computational efficiency and energy sustainability. For smart contracts, AI-driven approaches demonstrated high accuracy in vulnerability detection, with several models achieving detection rates above 90% [4], [14], [22]. Nevertheless, the widespread use of opaque deep learning architectures limited transparency and interpretability, posing risks for auditing and regulatory compliance in sensitive domains. In terms of ledger storage, AI-based compression techniques, such as auto encoders, initially reduced storage requirements by as much as 60% [11], [12]. Yet these benefits degraded over time and at scale, with one study reporting a 30% loss in compression efficiency during extended blockchain growth [13]. This suggests that while storage optimization is feasible, scalability remains a challenge. The analysis of sectoral applications reveals a strong dominance of finance, where eight out of ten studies focused on decentralized finance (DeFi) use cases [24]. However, these studies often relied on proprietary datasets, limiting reproducibility. In healthcare, federated learning models achieved promising diagnostic accuracy rates above 95% [6], [15], yet scalability was constrained, as some evaluations were based on fewer than 200 patients. Supply chain applications, while demonstrating improved logistics efficiency through RL-based IoT integration, remained heavily dependent on simulated environments, with five of six studies lacking realworld validation [16]. Beyond technical dimensions, the review highlights a broader reproducibility crisis. Only 12 of the included studies provided open-source code or publicly accessible datasets, while the majority (50) relied on proprietary data sources, restricting peer verification and extension. Similarly, ethical considerations were largely neglected, with 57 studies scoring ≤2/5 on quality assessment of normative and legal integration. This gap underscores the urgent need for actionable ethical frameworks and transparent research practices to support trustworthy AI–blockchain integration [27], [28]. V. DISCUSSION The discussion of findings highlights several critical themes emerging from the reviewed literature. A key limitation is the dominance of synthetic data and sectoral concentration, with finance and healthcare accounting for the majority of contributions. While these domains demonstrate tangible efficiency gains, such as improved liquidity prediction in DeFi and enhanced diagnostic accuracy in healthcare, the lack of real-world validation undermines generalizability. To address this gap, future studies should prioritize pilot projects and live blockchain deployments in underrepresented sectors such as supply chain logistics, agriculture, and energy, where practical challenges remain largely unexplored [16], [25], [26]. Another recurring issue is the superficial treatment of ethical and legal dimensions. Although several studies identified tensions between blockchain immutability and privacy regulations such as GDPR, few proposed actionable strategies for reconciliation. This poses significant legal risks, particularly in sensitive domains like healthcare and governance, where compliance failures could compromise adoption [8], [27]. Addressing these risks requires the integration of advanced privacy-preserving techniques, including zero-knowledge proofs (ZKPs) for selective data erasure and hybrid arbitration frameworks to manage liability in AI-driven contracts [10], [28]. The review also underscores the importance of decentralized AI approaches for preserving blockchain’s core ethos of distribution and transparency. Federated learning (FL), for instance, enables collaborative model training without centralizing sensitive data, thereby reducing the risks of bias concentration and power asymmetry in decentralized autonomous organizations (DAOs) [17]. However, these approaches must be complemented with robust governance structures to ensure equitable participation across nodes. Finally, emerging innovations such as self-healing contracts show potential to automate vulnerability detection and reduce manual auditing efforts by up to 40%. Yet, their adoption requires robust safeguards, including explainable AI (XAI) models that enhance interpretability and ensure regulatory compliance before such systems can be trusted in mission-critical environments. Table 6 highlights the major ethical and legal risks associated with AI–blockchain integration, including bias in decentralized governance, conflicts between GDPR and immutability, and liability gaps in automated contracts. The table also presents potential mitigation strategies, such as diversity-aware training datasets, ZKPs, and hybrid arbitration protocols. These strategies, while still largely conceptual, provide a roadmap for addressing the most pressing normative challenges in the field. TABLE 6: ETHICAL RISKS AND MITIGATION STRATEGIES Risk Sector Impact Proposed Solution Implementation Complexity G. Mandinyenya and Vusimuzi Malele, “Synthesizing the Future of AI-Blockchain Integration: A Pathway for Adaptive, Ethical, and Efficiency”, Latin-American Journal of Computing (LAJC), vol. 13, no. 1, 2026. Bias in AIDriven DAOs Finance, governance Diversityaware training datasets Moderate GDPR vs. Immutability Healthcare, public sector Zeroknowledge proofs for data erasure High Liability in Smart Contracts Legal, insurance Hybrid human-AI arbitration protocols Moderate Fig.4 below highlights the distribution of ethical and legal risks across severity levels, with GDPR conflicts and liability emerging as the most frequently cited high-impact. Fig.4 The distribution of ethical and legal risks across severity levels. One of the most pressing ethical challenges in AIblockchain integration concerns GDPR Compliance, particularly the tension between the “right to be forgotten” and blockchain’s inherent immutability. Recent proposals suggest that zero-knowledge proofs (ZKPs) can provide a pathway to reconciliation by enabling selective data erasure without compromising ledger integrity [8]. Another critical concern is liability in automated contracts, where responsibility for failures or disputes remains unclear. Hybrid human–AI arbitration frameworks have been proposed as a solution, ensuring accountability while retaining the efficiency benefits of automation [10]. For instance, in healthcare applications, GDPR-compliant blockchain systems could embed ZKPs to enable privacy-preserving patient record management, while in financial services, hybrid arbitration mechanisms could mitigate liability risks associated with DeFi transactions. A further dimension involves the challenge of transparency interpretability in AI-driven systems. Embedding explainable AI (XAI) within blockchain-based infrastructures offers a potential strategy to enhance trust, allowing stakeholders to audit decisions made by complex models without undermining efficiency or security [12], [18]. Fig.4. illustrates the AI-Blockchain interaction model (AIBIM), which highlights the layered synergy between consensus optimization, smart contract automation, and sector-specific applications. The model underscores how decentralized AI training and hybrid human-AI auditing can simultaneously strengthen resilience and preserve blockchain’s decentralization ethos. AI-BLOCKCHAIN INTERACTION MODEL (AIBIM) The Conceptual Model Fig. 4. AI Blockchain Interaction Model (AIBIM) Looking ahead, future work will focus on the empirical validation of the AI–Blockchain Interaction Model (AIBIM) through targeted case studies and prototype implementations. Such efforts will enable a practical assessment of the model’s scalability, security guarantees, and ethical robustness, thereby bridging the gap between conceptual design and realworld deployment. VI: LIMITATIONS While this review provides a comprehensive synthesis of AI–blockchain integration, several limitations must be acknowledged. First, the majority of the included studies (70%) relied on simulated environments, with only 12% validating their solutions on live blockchain networks [6], [15], [24]. This reliance on synthetic datasets limits the external validity of the findings and raises concerns about scalability in heterogeneous, real-world settings. Second, the reproducibility of results remains weak: only 15% of studies shared open-source code or datasets, creating barriers to peer validation and replication [13], [20]. This aligns with broader challenges in AI research, where proprietary data and closed implementations undermine transparency [27]. A further limitation is the sectoral bias observed in the literature. Finance and healthcare dominate existing contributions, while other critical industries such as energy, agriculture, and public governance remain underexplored [16], [25], [26]. This imbalance reduces the generalizability of insights and limits the applicability of proposed models to diverse domains. Finally, ethical and legal analyses across the reviewed studies were often theoretical rather than empirical, with 90% of papers lacking actionable frameworks to address bias, liability, or regulatory compliance [8], [10], [27]. Together, these limitations indicate the need for more diversified, reproducible, and empirically validated research to translate conceptual advances into deployable systems. VII: PRACTICAL IMPLICATIONS Despite these limitations, the findings of this review provide actionable insights for developers, regulators, and industry stakeholders. For developers, AI-driven consensus LATIN-AMERICAN JOURNAL OF COMPUTING (LAJC), Vol XIII, Issue 1, January - June 2026 . optimization and smart contract automation offer clear pathways to improve blockchain efficiency. Reinforcement learning, for instance, reduced consensus latency by up to 50% [1], [5], while NLP-based contract auditing improved vulnerability detection rates by over 40% [4], [14]. These innovations can be incorporated into prototype systems to enhance throughput and reduce manual verification. For regulators and policymakers, the results highlight the urgency of embedding privacy-preserving mechanisms such as zero-knowledge proofs (ZKPs) and federated learning into blockchain systems to reconcile immutability with GDPR’s “right to be forgotten” [8], [22]. Regulatory frameworks should evolve to account for liability in AI-driven contracts, particularly in decentralized finance (DeFi), where hybrid arbitration models could balance automation with accountability [10]. For industry practitioners, sector-specific findings point to immediate opportunities. In finance, AI-enhanced liquidity risk prediction models can strengthen DeFi resilience [24]. In healthcare, federated learning can enable GDPR-compliant medical data sharing while maintaining diagnostic accuracy [6], [15]. In supply chain management, reinforcement learning can optimize logistics efficiency, though pilot projects are needed to validate scalability [16]. By adopting the AI– Blockchain Interaction Model (AIBIM) proposed in this study, industries can systematically align technical innovations with governance and compliance requirements, accelerating the adoption of decentralized, intelligent infrastructures. VIII : FUTURE RESEARCH DIRECTIONS Building on the AI–Blockchain Interaction Model (AIBIM), which systematizes synergies across consensus, contract, and application layers, future research should prioritize translating conceptual advances into robust, deployable systems. A first priority is addressing the heavy reliance on simulated environments by developing real-world pilot deployments across finance, healthcare, supply chain, and underexplored sectors such as agriculture and energy [16], [25], [26]. Empirical case studies would provide the scalability evidence that is currently lacking. A second avenue involves advancing explainable AI (XAI) within blockchain contexts. While machine learning models improve smart contract auditing and vulnerability detection, their opacity undermines accountability. Embedding XAI techniques into blockchain systems could strengthen transparency, interpretability, and regulatory compliance [18], [27]. Third, reproducibility challenges must be resolved: only 15% of reviewed studies provided code or datasets, underscoring a critical barrier to validation and comparative analysis. Future work should therefore emphasize opensource benchmarking frameworks and standardized datasets to support peer validation and replication [13], [20]. Finally, ethical and legal frameworks require operationalization. Integrating zero-knowledge proofs (ZKPs), federated learning, and hybrid arbitration mechanisms could reconcile GDPR requirements with blockchain’s immutability, while also reducing liability risks [8], [22], [28]. Addressing these gaps will not only advance academic research but also accelerate practical deployment of AI– blockchain systems across finance, healthcare, supply chain, and emerging domains such as energy and agriculture, thereby bridging the gap between theoretical constructs and real-world decentralized infrastructures. VIIII: CONCLUSION This systematic review examined 28 peer-reviewed studies to assess how artificial intelligence (AI) is being applied to strengthen blockchain protocols, smart contracts, and ledger management. The evidence shows that AI-driven consensus mechanisms, such as reinforcement learning applied to Proof-of-Stake, can reduce latency by up to 50%, though at the cost of increased energy consumption [1], [5], [7]. Similarly, natural language processing has been used to generate and audit smart contracts, lowering vulnerabilities by as much as 40%, but raising concerns over transparency and auditability [4], [22]. Sectoral adoption has been most pronounced in finance and healthcare, while domains such as supply chain, agriculture, and energy remain underexplored [16], [25], [26]. Importantly, only a small proportion of the reviewed studies (12%) validated their approaches on live networks, highlighting the persistent gap between controlled experimentation and real-world deployment. Ethical and legal considerations re also limited. The immutability of blockchain continues to conflict with privacy requirements such as the GDPR’s “right to be forgotten,” with zero-knowledge proofs (ZKPs) and federated learning emerging as potential remedies [8], [20]. Yet, few studies propose concrete or testable frameworks to operationalize such solutions, leaving issues of liability, bias, and governance unresolved [10], [27]. The proposed AI–Blockchain Interaction Model (AIBIM) offers one pathway for addressing these challenges by systematizing synergies across consensus, contract, and application layers. It emphasizes decentralized AI training to preserve blockchain’s distributed ethos and hybrid human–AI auditing to enhance accountability at the contract layer. However, critical gaps remain. Reproducibility is weak, with only 15% of studies sharing open-source code or datasets. Ethical integration is insufficient, with 90% of studies lacking actionable mechanisms for fairness, liability, or accountability. Sectoral diversity is also lacking, with most work concentrated in finance and healthcare while public governance and energy remain underrepresented [26]. Future research should move beyond theoretical constructs by validating frameworks like AIBIM through prototypes, case studies, and benchmarking in live blockchain environments. At the same time, progress will require embedding explainability (XAI) and regulatory compliance at design level, ensuring that AI-enhanced blockchain systems are both technically robust and socially trustworthy [12], [28]. Achieving this will depend on interdisciplinary collaboration, particularly between computer science, law, and ethics, to ensure that AI–blockchain integration evolves into scalable, ethically aligned, and societally impactful solutions. REFERENCES