Artificial intelligence and ESG in resources-intensive industries: Reviewing the use of AI in fisheries, mining, plastics, and forestry

The study examines whether AI can improve the sustainability of global resource‑intensive industries, focusing on fisheries, mining, plastics, and forestry. It situates AI within ongoing CSR/ESG debates and the rise of corporate self‑governance in complex supply chains. While AI is touted for enhancing supply chain sustainability and responsible business conduct, the paper cautions that AI may accelerate extractivism, consumption, and waste, threatening environmental and social outcomes. The research addresses the gap in comparative analyses across multiple resource‑intensive sectors, outlining both potential benefits (e.g., data analysis for ESG risk detection, transparency, and operational efficiency) and risks (e.g., rebound effects, bias, job displacement, governance washing). The importance stems from these sectors' large socio-environmental footprints and centrality to the global economy and livelihoods.

The paper reviews AI’s roles and risks within ESG frameworks. OECD identifies key AI patterns used for ESG (pattern/anomaly detection, predictive analytics, decision support), alongside potential harms including bias and misuse. Multiple evaluative frameworks (e.g., ESG Digital and Green Index; Responsible AI frameworks integrating non-financial ESG metrics, AI ethics, and emerging regulation like the EU AI Act) assess AI’s sustainability and risk profiles. The literature highlights AI’s anti-corruption potential, lifecycle analysis support, and environmental applications, but flags digital pollution, bias, ethical gaps, and the risk of transferring responsibility to technology. Studies also note limited adoption of AI in CSR due to mixed perceptions and propose governance practices (e.g., CRaiO roles, transparent data policies, balanced economic-social objectives, incentives/regulation). The review underscores that most research evaluates AI system risks or direct sector risks rather than AI’s transformative impacts across sectoral sustainability, and critiques techno-optimist narratives that overlook consumption growth and systemic inequities.

Critical review approach focused on four resource‑intensive industries (fisheries, mining, plastics, forestry) selected for extractive nature, economic importance, and high socio-environmental impacts. Literature search via Google Scholar using broad keywords (AI and each industry, 2010–2025; with emphasis on 2015–March 2025). Articles were screened for alignment with ESG/sustainability. From 269 initial studies, 60 were selected: fisheries (15), mining (19), plastics (9), forestry (17). No new primary research was conducted; insights were informed by prior fieldwork in the DRC, Colombia, Ecuador, EU, South Africa, Tanzania, and the US. Findings per industry (Section 4) were synthesized into cross-industry trends (Section 5). This is a critical, not systematic, review.

- Overall: AI can yield direct, localized ESG improvements (efficiency, monitoring, predictive analytics, traceability), but indirect, system‑level effects (efficiency rebound, expanded extraction/production, inequality, governance washing) may outweigh benefits without robust policy guardrails. - Fisheries: AI supports stock assessments, vessel monitoring (AIS), IUU detection, catch monitoring, and traceability; can reduce waste and enforcement emissions via targeted patrols. Risks include overfishing through better fish detection, limited enforcement follow-through, hacking and surveillance risks, inequities disadvantaging small-scale fishers, privacy concerns, and job losses. - Mining (mineral industry): AI aids exploration (deposit targeting), construction and operations (route optimization, predictive maintenance, safety), and processing (fault detection, pollution monitoring), potentially reducing fuel use and localized pollution; mining contributes an estimated 4–7% of global GHG emissions. Risks include increased total extraction and localized pollution via expanded operations, pressures on Indigenous/agrarian lands, reduced labor demand (especially local), widening gaps with artisanal miners, high short-term equipment costs, and cybersecurity risks. - Plastics: AI enhances downstream waste management—robotic sorting, detection of macro/microplastics, waste modeling, recycling logistics, and traceability—supporting localized ESG gains. However, AI likely boosts upstream/midstream oil, gas, and petrochemical efficiency, reinforcing production growth and Jevons’ rebound, exacerbating pollution and GHG emissions. Globally, less than 10% of plastic waste has been recycled into new plastics since the 1950s; plastics production exceeds 450 million tonnes annually, with large-scale leakage to ecosystems. - Forestry: AI improves wildfire prediction/management, forest health monitoring (pests/diseases), biomass supply chain optimization, facility siting, species classification, and inventory tracking—benefiting environmental outcomes, rural livelihoods, and governance. Challenges include data scarcity/costs in remote areas, potential job displacement in rural/Indigenous communities, privacy/security issues in environmental/proprietary data, and interpretability concerns for regulatory decisions. - Cross-cutting governance: AI can enhance transparency, traceability, corruption detection, and regulatory alignment (e.g., EITI, due diligence laws), but risks governance washing, overreliance on technological fixes, and vulnerability to hacking. - Synthesis: Direct micro-level ESG benefits are evident; indirect macro-level risks (consumption-driven rebound, expanded extractivism, inequality, labor displacement, digital risks) threaten overall sustainability gains without comprehensive policy, regulation, and inclusive governance.

The findings address the core question by showing AI’s dual character: it can improve ESG performance through better data, prediction, and traceability, yet it can also entrench unsustainable patterns by scaling extraction/production and displacing labor. In fisheries, mining, plastics, and forestry, AI’s direct benefits arise from operational efficiency, monitoring, and risk mitigation; however, the indirect impacts—efficiency rebound (e.g., plastics), proliferation of mining projects via improved exploration, intensification of fishing effort, or job displacement in forestry—can undermine sustainability. Governance benefits (anti-corruption, transparency) depend on enforcement capacity and do not automatically translate into improved outcomes; weak enforcement, privacy concerns, and hacking erode effectiveness. The review underscores that AI is a supportive tool within a broader ESG toolkit, not a standalone solution; meaningful sustainability gains require policy guardrails, equitable access, and attention to biases and social impacts. Integrating EDI perspectives and addressing digital colonialism and data bias are central to ensuring AI’s ESG contributions are equitable and durable.

AI is increasingly embedded in ESG practices across fisheries, mining, plastics, and forestry, delivering data-driven insights, predictive capabilities, and traceability that can reduce localized harms and improve governance. Yet, without systemic reforms and strong regulation, AI risks amplifying extractivism, overconsumption, inequalities, and new digital vulnerabilities, thereby undermining sustainability. The paper contributes a cross-sector comparative review that identifies shared patterns of direct benefits and indirect risks, arguing for AI to be treated as one tool among many within comprehensive ESG strategies. Future research should: (1) assess sector-wide rebound effects and cumulative environmental and social impacts from AI deployment; (2) develop and test regulatory guardrails (e.g., under the EU AI Act) tailored to resource‑intensive sectors; (3) integrate EDI principles to mitigate bias and digital colonialism; (4) strengthen links between AI-generated insights and enforceable interventions; and (5) rigorously evaluate AI’s role across upstream, midstream, and downstream stages, especially in plastics and mining.

- Not a systematic review; critical review based on Google Scholar searches with broad keywords may introduce selection bias. - No new primary data collection; insights partially informed by prior fieldwork but reliant on published literature up to March 2025. - Focus limited to four industries; findings may not generalize to other resource‑intensive sectors (e.g., agriculture). - Heterogeneity of AI applications and contexts complicates cross-study comparability; some areas (e.g., AI traceability in mining transformation) remain emergent with limited empirical validation. - Tables in the literature occasionally emphasize benefits with incomplete reporting of challenges in some ESG dimensions, reflecting gaps in available studies. - Rapidly evolving AI and regulatory landscapes may outpace the review’s coverage.