Introduction

Global agriculture is at a decisive point: the need to reduce crop yield losses caused by pests to secure food security for an expected 9.2 billion people by 2040 must be balanced against the urgent need to reduce environmental harm [1]. Among the major contributors to ecological degradation is conventional pest management [2], which has heavily relied on synthetic insecticides (Fig 1). While synthetic pesticides have historically safeguarded yields, their overuse has resulted in soil degradation, water contamination, collapse of beneficial insect populations, and pest resistance [3,4]. Repeated exposure to the same active ingredients drives rapid evolutionary responses in pest populations, reducing efficacy and increasing dependence on newer, often more toxic, chemicals [4].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Illustration of the transformative impact of AI coupled with new molecular technology on insect pest monitoring toward resilient, sustainable agriculture.

Original figure (no AI used in its creation).

https://doi.org/10.1371/journal.pstr.0000216.g001

Insect monitoring for early pest detection is one of the keys for sustainable control of this harm. However, this strategy, used to rely on manual labor, is time-consuming and often results in delaying the apprehension of the current pressure status [5]. In this context, artificial intelligence (AI) provides a promising path to reimagining insect pest control. Combined with rapid and adaptable molecular tools [6], it enables accurate, reliable pest identification and determines whether a species acts as a vector (Fig 1). Leveraging computer vision, sensor networks, and predictive analytics, AI supports pest management strategies that are more targeted, timely, and environmentally conscious [7,8]. From real-time pest identification to outbreak forecasting, AI shifts pest control from reactive to proactive, offering a crucial step toward sustainability (Fig 1) [9].

AI: A transformative tool for insect pest management strategies

Recent advances in machine learning, have enabled accurate insect identification through image analysis (Box 1). More especially, convolutional neural networks have demonstrated better image-based insect classification potential and have been largely applied for classifying insect images [10]. As deep learning models, these algorithms can automatically extract relevant features in images, making them more accurate and more suitable for large dataset analysis, unlike traditional machine learning systems such as support vector machines which require manual feature engineering [11]. These models support timely decision-making and optimize broad-spectrum insecticides uses, thus preserving beneficial insect populations [12].

However, the effectiveness of AI in pest detection hinges on high-quality, diverse datasets. Current efforts such as AgriPest, Pest24, and IP102 have made significant progress, but only a fraction of the estimated 5,000 global pest species is represented [13–15]. To improve generalizability, training datasets must include images from different geographical regions, crop systems, and developmental stages. Public AI technologies now make it feasible to any citizen to process data on-site, even in low-connectivity settings. In this way, AI transforms data collection into a collaborative scientific endeavor that further improves detection models. As a result, farmers can access increasingly reliable real-time information, even in remote or resource-limited areas [12].

Integrating AI into sustainable agricultural systems

Beyond pest identification, AI supports a broader transition to sustainable agriculture (Box 2). Predictive models relying on an intelligent system can handle complicated information to better anticipate pest outbreaks using environmental variables like temperature, humidity, and crop phenology, allowing for well-timed interventions [13]. Indeed, unlike conventional decision support tools, AI-driven models don’t rely on predefined statistical assumptions, making them more flexible and accurate when dealing with complex and nonlinear variables [16]. AI also enables a better spatial efficiency of pest monitoring. Remote imaging using drones or unmanned aerial vehicles paves the way for tracking and monitoring insect pest infestations and migrations across larger spatial scales [17,18].

A better understanding of temporal and spatial dynamics of insect pests improves the use of conventional and alternative methods, such as releasing natural enemies. With AI-provided precision, pesticides could be applied only when and where needed, and IPM programs are strengthened (see Box 3). As a result, costs are lowered, pollution is limited, and the impact on global diversity is reduced.

Alongside with improvement of control strategies, AI can optimize cultural practices adjusting planting schedules, crop rotations, or tillage methods, all that based on insect pest behavior and climate data. This systems-level integration aligns with regenerative principles, reduces pest pressure and reliance on chemical controls [19].

Challenges and future directions

Despite its potential, AI adoption in agriculture faces technical and institutional barriers. One of the most pressing technical concerns is the access to standardized data management. To enable the widespread adoption of AI in agriculture, robust data management principles are essential for both performance and ethical integrity. Community standards, such as the FAIR Data Principles, enhance the ability of machines to find, access, integrate, and reuse scientific data [20]. However, FAIR does not address relationships, power dynamics, or the historical contexts in which data are produced and used [21]. To address these dimensions, FAIR can be complemented by frameworks that emphasize responsible and ethical data practices, such as the CARE Principles.

Datasets disproportionately sample large, industrial farms, underrepresenting smallholder and remote-farm contexts, especially in developing countries, where these producers account for a major share of agricultural production [22]. Moreover, tools developed in academic, or industry often fail to meet the needs of growers, especially smallholder farmers. Interfaces must be intuitive, multilingual, and compatible with on-farm realities to ensure widespread use (see Box 3).

Enabling policy environments are also essential. Current regulations seldom account for AI-based tools in agriculture, leaving gaps in accountability, data protection, and technology validation. Public investment and governance frameworks must catch up to support safe and equitable deployment [23].

Future research should prioritize:

• Building open, diverse datasets representative of the diversity of agricultural contexts.
• Improving model robustness to variable lighting, occlusion, and mixed insect pest populations by including field-based dataset during training.
• Keep existing solutions up to date with current data and improve the accuracy of detection models while reducing the cost of AI-dedicated hardware. Promoting farmer–scientist collaborations in model co-design.

Conclusion

AI holds tremendous promise to shift insect pest control from a chemical-intensive model to one grounded in ecological intelligence and precision (Fig 1). When embedded within broader sustainable agriculture frameworks, AI has the potential to transform not just how insect pests are managed, but how food is grown, ecosystems are restored, and resilience is built into the future of farming. To realize this potential equitably, it is essential that AI tools are co-developed with farmers, researchers, and policymakers to ensure relevance, accessibility, and long-term impact.

Finally, academics can help turn promise into practice by co-developing AI tools with growers and agencies; curating open, representative datasets; running on-farm validation and benchmarking; and producing standards, training, and extension materials. These actions address key technical (data gaps/bias, interoperability, field robustness), regulatory (evidence, privacy/data rights, liability/approval), and social (trust, usability, language/connectivity, cost) barriers, while giving policymakers actionable evidence. A participatory, open, standards-driven agenda will accelerate safe, equitable AI adoption in insect pest monitoring across farm sizes and regions.

Box 1 Progress of AI-based image identification techniques in key insect orders relevant to agriculture.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

https://doi.org/10.1371/journal.pstr.0000216.t001