Personally identifiable information (PII) detection is foundational to privacy-preserving analytics, compliant data sharing, and safe model training. Traditional rule-based and supervised named entity recognition (NER) pipelines excel on templated fields but are brittle under domain drift and long-tail variants, motivating large language models (LLMs) that generalize via in-context learning. Despite rapid LLM progress, we lack systematic and cross-domain evaluations that reflect redaction-centric error costs and practical deployment constraints. This paper presents an LLM-centric benchmark of PII detection across three textual domains and introduces a carefully annotated multimodal PII dataset to extend beyond text-only settings. Our experiments examine zero-shot and few-shot prompting, quantify prompt sensitivity, and characterize privacy-relevant error modes. We further assess a privacy-oriented, cost-aware baseline that combines a small open-source model with a multi-agent retrieval-augmented workflow, analyzing performance and feasibility for resource-constrained environments. Results highlight domain and modality effects, consistent gains from few-shot prompting (with saturation effects), and precision–recall trade-offs that vary across tasks and models. The proposed agentic retrieval augmented generation (RAG) pipeline approaches, and in some cases exceed the performance of large proprietary models while improving privacy posture and operational cost. The prompts and the multimodal dataset are made publicly available to support reproducibility and future benchmarking.

Avian influenza virus (AIV) continues to pose serious risks to animal and public health. Understanding its spread requires integrating ecological, agricultural, and human information. Quantitative models provide a practical way to represent these interactions, evaluate multiple risk factors, and generate spatial insights that support early detection and control. In recent years, advances in data availability and computational methods have increased the use of these models for AIV risk assessment.This review mapped how quantitative modelling has been applied to AIV risk assessment in recent years. Following PRISMA-ScR guidelines, we searched PubMed, Web of Science, and ProQuest for studies published between January 2020 and March 2025, identifying 34 eligible studies. Five model categories were identified: Logistic Regression-based Models, Generalized Linear Models (GLMs), Machine Learning (ML), Multi-Criteria Decision Analysis (MCDA), and Exploratory Statistical Models. Logistic regression and GLMs quantified associations between outbreaks and risk factors, while ML models focused on predictive mapping. MCDA combined expert weighting with spatial data to generate composite risk maps, and exploratory analyses examined spatial and temporal outbreak patterns.Despite methodological progress, inconsistencies remain in data preparation and validation, limiting comparability across studies. Clearer, harmonized workflows are needed to improve reproducibility and support translation into surveillance tools globally.

Adopting artificial intelligence (AI) tools and technologies may be favourable for the agriculture sector in addressing food security and sustainability. As AI gains popularity, there are clear benefits to using AI in the agriculture sector. However, without effective governance, there are risks and challenges that can limit the benefits that can be obtained from using AI. For instance, ineffective policies that fail to outline liability in the event of adverse outcomes, such as crop loss resulting from ineffective decision-making by an AI technology or tool. We present a definition of AI governance in agriculture, in addition to a novel taxonomy which is comprised of nine principles: inclusivity & education, trustworthiness, custodianship & liability, law, sustainable & ethical development, transparency, privacy & security, bias, and data quality. Following a scoping review methodology, we conduct an analysis of the existing literature on AI governance in agriculture through the lens of the nine principles. Additionally, we uncover factors which support the principles of AI governance outlined in the discussion. Furthermore, identifying research gaps provides a clear road map for future directions, which can support researchers, policymakers, and practitioners. This study supports the ongoing development of AI governance and sheds light on the critical need for tailored governance for the agriculture sector.

Highly Pathogenic Avian Influenza (HPAI) poses a severe biological and economic threat to the Canadian poultry industry. The transient and day-to-day meteorological and regional triggers remain poorly quantified due to spatial confounding in traditional epidemiological models. To isolate these triggers, this study utilizes a time-stratified case-crossover design, integrating a veterinary surveillance dataset of infected premises (2022–2024) with ERA5 meteorological reanalysis data. We employ the conditional difference method optimized with Ridge (L2) regularization. The national model demonstrated discriminative ability (ROC-AUC = 0.978; Brier score = 0.060). The analysis identified local proximity to an active outbreak as the strongest statistical predictor, with patterns consistent with wind-mediated transmission amplified by high wind speeds and relative humidity. A provincial-level stratification revealed that transmission mechanics are geographically distinct. In dense agricultural zones, extreme farm proximity dictates an effect that overpowers ambient weather factors. In dispersed geography of the Prairies, meteorological factors dominate infection triggers. These findings suggest that mitigation interventions must be tailored to regional factors.

Avian Influenza Viruses (AIV) pose ongoing risks to the poultry industry, wildlife, and public health, with highly pathogenic avian influenza (HPAI) outbreaks causing substantial economic losses and ecological impacts. Early identification of high-risk areas is critical for disease management, yet current approaches relying on manually engineered tabular datasets are often limited in capturing complex spatial dependencies. In addition, the sporadic nature of AIV outbreaks results in data scarcity and class imbalance, which limits the effective use of deep learning models trained from scratch. To address these challenges, this study presents the Regional AIV Vision-based Encoder Network (RAVEN), a raster-based deep learning model for predicting AIV outbreak risk in the near future using multi-source geospatial data. RAVEN leverages transfer learning with a pretrained foundation model (Vision Transformer) to address limited labeled data. This approach is evaluated using a nationwide grid-based dataset covering Canada’s land territory. Across the domestic and wild bird datasets, RAVEN achieved F1 scores above 90% and 95%, respectively, consistently outperforming classical machine learning baselines. Analysis showed that the model prioritizes ecologically significant features, such as specific wetland types, weather variables supporting viral stability, and high-density poultry sectors. These results highlight the potential of applying computer vision techniques to disease surveillance and suggest that foundation models can enhance early warning systems for zoonotic diseases.

CONTEXT The digitalisation of agriculture through Internet of Things (IoT) technologies is promoted as a driver of efficiency and sustainability, yet adoption remains uneven, particularly where farmers face barriers of skills, trust, and infrastructure. OBJECTIVE This study explores how Canadian farmers, industry representatives, and government officials construct discourses around IoT adoption, and how these shape opportunities, constraints, and legitimacy. METHODS We conducted 47 semi-structured interviews with farmers, industry actors, and government officials. Thematic analysis identified 12 themes, organised along two continua: individual versus structural opportunities and rational versus cultural elements. The analysis was guided by a dual lens framework of opportunity structures and loose coupling. RESULTS AND CONCLUSIONS Adoption depended on both individual capacities (training, simplicity, digital literacy) and structural conditions (labour, routines, connectivity). Legitimacy was negotiated through rational evaluations of utility and risk and cultural considerations such as values, ethics, and identity. Adoption unfolded through loosely coupled discourses, producing incremental and uneven rather than uniform diffusion. SIGNIFICANCE The study advances debates on agricultural digitalisation by showing how adoption is experienced through everyday practices, negotiated legitimacy, and loosely coupled diffusion. Beyond Canada, the framework offers a transferable lens for analysing contested adoption in agriculture and other socio-technical systems.

Explainable artificial intelligence (XAI) plays a crucial role in mitigating the risks associated with the non-transparency of black-box artificial intelligence (AI) systems. However, despite its advantages, XAI methods have been shown to expose the privacy of individuals whose data are used to train or query the underlying models. Prior research has demonstrated privacy attacks that exploit explanations to infer sensitive personal information of individuals. At present, there is a lack of effective defenses against such privacy attacks targeting explanations, particularly when vulnerable XAI techniques are deployed in production environments or used in machine learning as a service systems. To address this gap, this study investigates the use of privacy enhancing technologies (PETs) as a defense mechanism against attribute inference attacks on explanations generated by feature-based XAI methods. We empirically evaluate three types of PETs, i.e., synthetic training data, differentially private training and noise addition, across two categories of feature-based XAI. Our findings reveal varying levels of effectiveness among the mitigation strategies, as well as trade-offs between privacy, utility and system performance. In the best scenario, integrating PETs into the explanation process reduced attack success by 49.47% while preserving model utility and explanation quality. Based on our evaluation, we propose strategies for effectively integrating PETs into XAI to maximize privacy protection and minimize the risk of sensitive information leakage.

Black-box Artificial Intelligence (AI) systems have achieved state-of-the-art accuracy in many problem domains in recent years. However, the lack of transparency of these systems is a bottleneck in their usage in high-risk applications which make automated decisions on individuals. Trustworthy AI proposes principles such as reliability, validity, privacy, fairness, and explainability among others, to mitigate risks from large-scale AI deployments in such domains. Explainable AI (XAI) is a technique of providing insights into the decision-making process of black-box systems thus enabling transparency. It plays a crucial role in communicating the rationale of automated decisions to relevant stakeholders. Though explainability is a highly desirable requirement, recent research has determined that explanations can introduce new privacy risks in AI systems. Researchers have demonstrated different types of privacy attacks on XAI deployed in production and cloud systems. Despite these risks, currently there is a lack of research into defenses for known privacy attacks in XAI. In this article, we contribute to this gap by proposing a defense mechanism for attribute inference attack on feature-based XAI. We empirically evaluate a well-known privacy preservation technique, namely, additive noise, and show its impact on privacy, explainability and utility. Our findings indicate that additive noise enables privacy while achieving faithful explanations and without compromising model utility.

The agriculture and food (agri-food) sector faces rising global concerns about its sustainability and resilience to climate events. Thus, new solutions are needed to ensure environmental and food security. Artificial Intelligence (AI) offers inventive solutions to improve agricultural and food production practices. Generative AI methods, such as generative adversarial networks (GANs), variational autoencoders, and large language models (LLMs), add to the transformative process initiated by AI and expert systems in agricultural and food-related practices to enhance productivity, sustainability, and resilience. This study categorizes generative AI approaches and their capabilities in agri-food systems and provides a comprehensive review of the current landscape of generative AI applications in the sector. It discusses the impact of these technologies on enhancing agricultural productivity, food quality, and safety, as well as sustainability, presenting potential use cases like combatting climate change and foodborne disease modeling that highlight the practical applications and benefits of generative AI in agri-food. Furthermore, it addresses the ethical implications of deploying generative AI, including privacy, security, reliability, and unbiased decision-making.

As organizations continue to embrace the use of artificial intelligence (AI) systems, it is crucial to ensure that these AI systems can be trusted. However, there is still a significant gap between research on trustworthy AI and its implementation in real-world applications. To address this issue, we sought to explore the perspectives of AI developers and the challenges they face in creating trustworthy AI systems. This exploratory study involved conducting interviews with 19 AI developers. We identified key challenges faced by AI developers due to the immature state of trustworthy AI, inconsistent global regulatory landscape, a lack of standardized definitions of key concepts, limited tools and standards for practical implementation in organizations. This paper provides recommendations for organizations to invest in trustworthy AI processes and practices, this includes building a foundation for trustworthy AI specific to their organization, adopting an organizational approach to trustworthy AI culture, and providing proper data infrastructures to support AI developers in creating trustworthy AI systems. By investing in trustworthy AI practices, organizations can prepare for evolving regulations and ensure that their AI systems are reliable and trustworthy.

Large language models (LLMs) and generative artificial intelligence (AI) have demonstrated notable capabilities, achieving human-level performance in intelligent tasks like medical exams. Despite the introduction of extensive LLM evaluations and benchmarks in disciplines like education, software development, and general intelligence, a privacy-centric perspective remains underexplored in the literature. We introduce Priv-IQ, a comprehensive multimodal benchmark designed to measure LLM performance across diverse privacy tasks. Priv-IQ measures privacy intelligence by defining eight competencies, including visual privacy, multilingual capabilities, and knowledge of privacy law. We conduct a comparative study evaluating seven prominent LLMs, such as GPT, Claude, and Gemini, on the Priv-IQ benchmark. Results indicate that although GPT-4o performs relatively well across several competencies with an overall score of 77.7%, there is room for significant improvements in capabilities like multilingual understanding. Additionally, we present an LLM-based evaluator to quantify model performance on Priv-IQ. Through a case study and statistical analysis, we demonstrate that the evaluator’s performance closely correlates with human scoring.

As AI systems continue to increase their capabilities of performing human tasks there is a growing need to understand how the AI system determined its decisions. Interpretability is a concept in trustworthy AI research that is focused on understanding of the inner workings and decisions that come from the AI system. Our previous research revealed that AI developers lack consistent approaches or tools for implementing interpretability. There has been substantial theoretical interpretability research, yet the development of practical approaches in the form of tools to assist AI developers on interpretability remains underexplored in the research. This paper develops a taxonomy of AI interpretability techniques based on an analysis of the research literature using the survey of surveys method across 70 papers, examining 30 papers from 2019 to 2023. This paper develops a hierarchical taxonomy at a lower-level of abstraction to present and build upon relevant research literature. This paper also provides an approach as a practical tool in the form of decision trees to help AI developers identify and categorize interpretability techniques based on applicability and characteristics. Unlike existing interpretability taxonomies, this study introduces AI developer-oriented sample decision trees to aid in operationalizing the selection of interpretability techniques. This bridges theoretical research with practical implementation for AI developers. Preliminary testing and validation with AI developers was conducted to ensure applicability of the proposed taxonomy and decision trees. Preliminary validation demonstrated conceptual clarity and practical relevance, forming a foundation for future large-scale evaluation. This research contributes to further bridge the gap between AI systems and ensuring the practical implementation of interpretability.

Avian Influenza Virus (AIV) poses significant threats to the poultry industry, humans, domestic animals, and wildlife health worldwide. Monitoring this infectious disease is important for rapid and effective response to potential outbreaks. Conventional avian influenza surveillance systems have exhibited limitations in providing timely alerts for potential outbreaks. This study aimed to examine the idea of using online activity on social media, and Google searches to improve the identification of AIV in the early stage of an outbreak in a region. To this end, to evaluate the feasibility of this approach, we collected historical data on online user activities from X (formerly known as Twitter) and Google Trends and assessed the statistical correlation of activities in a region with the AIV outbreak officially reported case numbers. In order to mitigate the effect of the noisy content on the outbreak identification process, large language models were utilized to filter out the relevant online activity on X that could be indicative of an outbreak. Additionally, we conducted trend analysis on the selected internet-based data sources in terms of their timeliness and statistical significance in identifying AIV outbreaks. Moreover, we performed an ablation study using autoregsressive forecasting models to identify the contribution of X and Google Trends in predicting AIV outbreaks. The experimental findings illustrate that online activity on social media and search engine trends can detect avian influenza outbreaks, providing alerts earlier compared to official reports. This study suggests that real-time analysis of social media outlets and Google search trends can be used in avian influenza outbreak early warning systems, supporting epidemiologists and animal health professionals in informed decision-making.

The exponential growth of social media usage has raised significant privacy concerns, especially regarding adversarial authorship attribution. Social media texts pose distinctive privacy risks for authorship attribution due to their informal, high-variance style. We evaluate Privacy by Translation (PbT) and a hybrid that applies Differential Privacy during classifier training (PbT+DP) to mitigate authorship attribution while preserving utility for social-media mental-health classification. PbT translates posts into semantically equivalent text that obscures authorial style. PbT is formulated in a two-step process (translation and learning), and we evaluate the resulting privacy-utility trade-off. We instantiate PbT with Spanish and Arabic and compare against paraphrasing, perturbation, and DP-only baselines, reporting translation fidelity and model explainability. Our findings reveal that a hybrid approach combining DP with PbT outperforms standalone PbT, DP, and existing methods in balancing privacy and utility in social media mental health classification. The hybrid method reduces adversarial authorship classification F1-score by 89.1% and maintains an efficient privacy-utility ratio of 0.40. We also investigate the transformation of linguistic features, including stylistic, syntactic, and semantic elements, to understand their roles in authorship attribution.

Privacy attacks on machine learning (ML) models pose significant risks to individuals whose personal data is used for training or querying these models. Although concerns about the potential exposure of sensitive information through ML models continue to grow, existing safeguard mechanisms primarily focus on security threats, often neglecting privacy risks. In this paper, we examine existing tools to assess privacy risks of ML models and provide an overview of various privacy attacks and defense strategies. Given the lack of a comprehensive framework for assessing privacy vulnerabilities, we propose a generic framework for evaluating the privacy of ML systems and establish a set of tailored evaluation metrics for different types of privacy attacks. In addition, we develop a dedicated testbed to implement our framework and present experimental results that demonstrate the impact of various privacy attacks on different ML models.

The emergence of smartphones and internet accessibility around the globe have enabled billions of people to be connected to the digital world. Due to the popularity of instant messaging applications and social media, a large quantity of personal data is in text format, and processing text data in a privacy-preserving manner poses unique challenges. While existing reviews focus on privacy concerns from specific algorithmic perspectives or target only a particular domain, such as healthcare or smart metering, they fail to provide a comprehensive view that addresses the multi-layered privacy risks inherent to text data processing. Existing works often limit their scope to specialized solutions like differential privacy, anonymization, or federated learning, neglecting a broader spectrum of challenges. To fill this gap, we present a comprehensive review of privacy-enhancing solutions for text data processing in the present literature and classify the works into six categories of privacy risks: (i) unintentional memorability, (ii) membership inference, (iii) exposure and re-identification, (iv) language models and word embeddings, (v) authorship attribution, and (vi) collaborative processing. We then analyze existing privacy-enhancing solutions for text data by considering the aforementioned privacy risks. Finally, we identified several research gaps, including the need for comprehensive privacy metrics, explainable algorithms, and privacy in social media analytics.

The ongoing avian influenza outbreaks have had significant implications for the global poultry industry in addition to a wide range of wild birds and mammals. To enhance our understanding of public perceptions and reactions during such outbreaks, the present study examined social media discourse surrounding avian influenza on X (formerly known as Twitter). By employing advanced large language models, including DistilBERT for post filtering (average 89.5% accuracy via 5-fold cross-validation) along with Mixtral-8x7B, BERTopic, and RoBERTa for sentiment and topic/user analysis, this research categorizes the discussions and sentiments expressed by users over time. Our analysis focused on three aspects: main topics, sentiment, and temporal patterns of user engagement surrounding avian influenza outbreaks. Sentiment analysis revealed that a majority of posts related to economic impact (81.2%), wildlife (71.7%), and human cases (67.9%) expressed negative sentiment. Through topic modeling, prevalent topics of concern were identified in discussions, including concerns about transmission to humans and mammals, as well as issues related to food security and food prices. Additionally, the analysis of user engagement patterns showed distinct categories of users and highlighted the contributions of top users in shaping the discourse. Emotion analysis showed that over 80% of posts on major topics conveyed emotions such as anger, sadness, and fear, especially during periods of high case reports. The present study underscores the potential of social media analysis to understand public reactions to avian influenza outbreaks and to facilitate effective responses to public concerns and needs.

As AI-driven solutions become increasingly common, end users of those systems continue to lack transparency in understanding the AI system’s functionality leading to a lack of trust, and AI systems not reaching their full potential. Our research aims to address this research gap by developing a novel sample transparency card. This transparency card is grounded in key transparency requirements from the prominent AI legislation (the European Union’s (EU) AI Act), the international standards (national institute of standards and technology (NIST) and international organization for standardization (ISO)), and public transparency principles from 25 global organizations. The research follows a 2-phase research approach. First, we analyze the transparency requirements to gain a greater understanding of the legislative and organizational requirements. Second, based on transparency requirements, create common categories and develop a proposed sample transparency card. This research offers a novel contribution by developing a sample transparency card that can be seen by the AI end users and aligns with legislative acts and frameworks, promoting trust in AI systems.