The review article investigates the role of AI in improving the sustainability of resource-intensive industries. It addresses the growing techno-optimist literature surrounding AI's potential to enhance ESG practices within global supply chains, acknowledging both the hopes and real-world limitations. The research question is whether AI can genuinely improve the sustainability of these industries, considering both intended and unintended consequences. The article focuses on four case studies: fisheries, mining, plastics, and forestry, chosen for their extractive nature, reliance on natural resources, and significant economic and social impacts. The review bridges a gap in the literature by offering a comparative analysis of AI-powered ESG implementation across multiple sectors, highlighting commonalities and differences. The authors acknowledge the importance of data analysis in ESG and recognize AI's potential to process vast datasets. However, they caution that AI's widespread use to optimize consumption, without addressing overconsumption, might lead to unsustainable global practices. The introduction sets the stage by discussing the historical context of Corporate Social Responsibility (CSR) and ESG, emphasizing the role of corporations in improving livelihoods and mitigating inequalities within global supply chains. It also highlights the industry's adoption of self-governance structures to preempt government regulations and the integration of AI systems to support these initiatives. Finally, it emphasizes AI's potential to be a tool among many, rather than a sole solution for sustainability issues. The authors point to the need for a critical approach recognizing AI's potential to exacerbate unsustainable practices.

The paper reviews existing literature (2015-March 2025) on AI applications in fisheries, mining, plastics, and forestry, focusing on ESG criteria. The literature review identifies several key themes, including the different ways AI systems (as categorized by the OECD) can be used for ESG purposes. The review examines the various frameworks used to evaluate ESG issues related to AI, highlighting the importance of addressing ethical challenges, responsible AI, and the risk of biases. The authors note that current models tend to focus more on the risks and sustainability of AI systems themselves and their direct relation to sectors rather than the transformative potential of AI on the sustainability of the sectors. The literature review highlights that while AI holds potential in mitigating corruption and improving environmental outcomes, adoption remains limited due to mixed perspectives on risks and benefits. The discussion highlights existing frameworks for responsible AI, but notes a lack of comparative assessments across resource-intensive industries, which this review seeks to address.

The article employs a critical review approach to analyze the literature on AI-powered ESG implementation in four industries. The authors identified publications exploring AI applications in each of the four resource-intensive industries with a focus on incorporating ESG criteria. The review encompasses publications from 2015 to March 2025. Initially, they identified 269 studies using Google Scholar searches with keywords such as "AI" and the specific industry. Publications were selected based on their relevance to ESG and/or sustainability. A critical assessment of abstracts narrowed the selection to 60 studies: 15 for fisheries, 19 for mining, 9 for plastics, and 17 for forestry. The authors analyzed the selected articles to identify commonalities and differences in AI implementation across the four industries. This approach fills a gap in the literature by providing a comparative analysis of AI's use in these resource-intensive sectors. The review uses the key ESG issues classified by the CFA Institute to structure the analysis for each case study. The authors acknowledge that their approach is a critical review and not a systematic review but aims to assess the challenges and benefits of AI in the chosen industries, leveraging their fieldwork experience in various locations.

The case studies reveal both the potential benefits and significant challenges of AI in each sector: **Fisheries:** AI can improve stock assessments, monitor fishing practices, detect illegal activities, enhance traceability, and reduce abuses. However, it may increase overfishing, exacerbate inequalities, and raise privacy concerns. **Mining:** AI can optimize resource extraction, reduce GHG emissions, improve safety, and mitigate pollution. But, it can also lead to increased overall mining activity, disproportionately affecting communities and possibly increase inequalities between industrial and artisanal mining. AI can be used to predict potential incidents in mining such as water and soil contamination **Plastics:** AI can improve waste management, recycling rates, and the detection of plastic pollution. However, it could also incentivize unsustainable production of plastics through increased efficiency, not addressing the root causes of plastic pollution (overproduction). **Forestry:** AI enhances forest management, wildfire prediction, and disease detection, improving efficiency and sustainability. However, it can lead to job displacement in rural and indigenous communities, and raises data privacy and security concerns. Across all four industries, AI has the potential for positive direct impacts in improving sustainability, but also has the potential for significant negative indirect impacts such as increasing overall resource extraction, consumption, and pollution if not implemented cautiously.

The findings highlight the complex and often paradoxical role of AI in promoting ESG. While AI can improve efficiency and transparency, leading to direct positive environmental and social impacts, it can also indirectly exacerbate unsustainable practices and inequalities. This paradox underscores the importance of carefully considering both direct and indirect consequences when deploying AI in resource-intensive industries. The discussion emphasizes the limitations of AI as a standalone solution for achieving sustainability. Instead, AI should be viewed as a tool within a broader strategy that addresses systemic issues like overconsumption and unequal power dynamics. The authors emphasize the need for policy interventions, regulatory frameworks, and ethical guidelines to guide the development and implementation of AI for ESG purposes. They stress the necessity of moving beyond incremental improvements toward transformative change that challenges existing power structures and promotes equitable outcomes.

The study concludes that while AI offers significant potential to improve ESG in resource-intensive industries, its implementation requires careful consideration of both direct and indirect impacts. AI tools can enhance efficiency, transparency, and accountability, leading to positive outcomes in areas like pollution reduction, waste management, and improved labor practices. However, AI's potential to increase resource extraction and consumption rates could outweigh these benefits without robust policies to address overconsumption and ensure equitable distribution of the technology’s benefits. The authors emphasize the need for a holistic approach that integrates AI with broader policy interventions to promote genuine sustainability and address the structural challenges inherent in late-stage capitalism. Future research should focus on evaluating the long-term impacts of AI on ESG, exploring the development of ethical guidelines and regulatory frameworks, and addressing the potential for AI to exacerbate existing inequalities.

The study's limitations include its reliance on a critical review of existing literature, rather than primary research. The focus on four specific industries may limit the generalizability of the findings to other sectors. The selection of literature might inadvertently reflect biases in existing research. The timeframe of the review might not capture the most recent advancements in AI technologies or policies, and the indirect impacts of AI in these sectors are difficult to capture completely.