Vulnerability Of Systems Research Articles

Optimizing the photovoltaic-assisted electric bus network with rooftop energy supply

As a clean and renewable resource, solar energy has demonstrated its potential to alleviate the energy vulnerability and grid strain for electric bus systems. In this study, we investigate the optimal design of an electric bus network in which rooftop solar panels are equipped to provide en-route photovoltaic assistance. A continuous-based model is proposed to optimize critical network design variables, including time-varying headway, stop spacing, and deployment of depot chargers. Introducing rooftop solar photovoltaics can significantly reduce range anxiety for electric buses. Compared with conventional electric bus systems, 4.2%–7.8 % daily in-operation time can be extended, while at most 5.7 % fleet size can also be saved per day. The proposed system demonstrates much more advantages in regions where hourly solar irradiance exceeds 1000 W/m2. Sensitivity tests indicate that (1) with a 10 % improvement in energy conversion efficiency, 1.2 % daily fleet size and 7.5 % depot chargers can be reduced; (2) every 50-kWh increase in bus battery capacity would bring a 1.8 % increment in system cost. Increasing the installation area of solar panels can help reduce battery capacity and lower the system cost. A real-world-based case in Panjin City, China is further designed to validate the application of the proposed system.

Systematic survey on artificial intelligence based mobile crowd sensing and sourcing solutions: Applications and security challenges

Mobile Crowd Sensing/Souring (MCS) is a novel sensing approach that leverages the collective participation of users and their mobile devices to collect sensing data. As large volumes of data get stored and processed by the MCS platform, Artificial Intelligence (AI) techniques are being deployed to make informed decisions that help optimize the system performance. Despite their effectiveness in solving many of the challenges, incorporating AI models in the system introduces many concerns, which could adversely affect its performance. This includes exploiting the vulnerabilities of the models by an adversary to manipulate the data and cause harm to the system. Adversarial Machine Learning (AML) is a field of research that studies attacks and defences against machine learning models. In this study, we conduct a systematic literature review to comprehensively analyze state-of-the-art works that address various aspects of AI-based MCS systems. The review focuses mainly on the applications of AI in different components of MCS, including task allocation and data aggregation, to improve its performance and enhance its security. This work also proposes a novel classification framework that can be adapted to compare works in this domain. This framework can help study AML in the context of MCS, as it facilitates identifying the attack surfaces that adversaries can exploit, and hence highlights the potential vulnerabilities of AI-based MCS systems to adversarial attacks, motivating future research to focus on designing resilient systems.

APPROACH TO THE ANALYSIS OF SUCCESS CRITERIA AS A COMPONENT OF PROBABILISTIC PHYSICAL PROTECTION ANALYSIS OF NUCLEAR FACILITIES

An approach to adapting the methodology for analyzing success criteria from traditional probabilistic safety analysis (PSA) of nuclear power plants (NPPs) to probabilistic physical protection analysis (PPA) of nuclear facilities is demonstrated using the hypothetical nuclear facility SNRI as an example. Generally speaking, using the traditional PSA methodology for assessing the physical protection of nuclear facilities allows for the identification of new vulnerabilities in the physical protection systems of nuclear facilities, the development of recommendations to enhance physical protection efficiency, the performance of more substantiated cost-benefit analyses, and more.

Identifying vulnerabilities in essential health services: Analysing the effects of system shocks on childhood vaccination delivery in Lebanon

Shocks effects are under-theorised in the growing literature on health system resilience. Existing work has focused on the effects of single shocks on discrete elements within the health system, typically at national level. Using qualitative system dynamics, we explored how effects of multiple shocks interacted across system levels and combined with existing vulnerabilities to produce effects on essential health services delivery, through the prism of a case study on childhood vaccination in Lebanon. Lebanon has experienced a series of shocks in recent years, including large-scale refugee arrivals from neighbouring Syria, the COVID-19 pandemic and a political-economic crisis. We developed a causal loop diagram (CLD) to explore the effects of each shock individually, and in combination. The CLD was developed and validated using qualitative data from interviews with 38 stakeholders working in Lebanon's vaccination delivery system, in roles ranging from national level policy to facility-level service delivery, conducted between February 2020 and January 2022.We found that each of the shocks had different effects on service demand- and supply-side dynamics. These effects cascaded from national through to local levels. Both Syrian refugee movement and the COVID-19 pandemic primarily exposed vulnerabilities in service demand, mainly through slowly emerging knock-on effects on vaccination uptake behaviour among host communities, and fear of contracting infection in crowded health facilities respectively. The economic crisis exposed wider system vulnerabilities, including demand for vaccination as household income collapsed, and supply-side effects such as reduced clinic time for vaccination, declining workforce retention, and reduced availability of viable vaccine doses, among others. Finally, important pathways of interaction between shocks were identified, particularly affecting the balance between demand for vaccination through publicly supported facilities and private clinics. Future research should incorporate dynamic approaches to identifying within-system vulnerabilities and their potential impacts under different scenarios, as a precursor to improved resilience measurement, system preparedness, and intervention targeting.

Climate change loss and damage from droughts: key insights from Fiji’s sugar industry

ABSTRACT Climate change loss and damage (L&D) due to both sudden and slow-onset events are a growing concern for vulnerable countries. Over the last decade, agricultural communities in Fiji have suffered a range of serious L&D from severe drought events. Through a grounded theoretical lens, qualitative research was used to gain in-depth insights into L&D from droughts in the Fijian sugar industry. In-depth semi-structured interviews (n = 68) were conducted in two Indo-Fijian sugarcane communities, Barotu and Toko settlements in Western Viti Levu, Fiji, and with key stakeholders from government ministries, academia, and climate change experts at the national level. Purposive sampling was initially used to identify smallholder sugarcane farmers and key stakeholders. Theoretical sampling further identified key stakeholders for the interview. Despite implementing various adaptation measures, the Fijian sugar industry has suffered severe L&D, including non-economic L&D (NELD) and associated cascading effects. L&D included reduced yields, loss of crops, and reduced income. NELD included deterioration of mental, physical, and emotional health, loss of hope, and uncertainty. Cascading effects included increased food insecurity risks and impacts on children’s education. Key policy interventions are recommended, such as removing adaptation constraints, developing drought risk profiles and early warning systems, investing in livelihood diversification, and engaging in new markets to facilitate social and ecological transformations that will promote livelihood resilience. Key Policy Highlights: Climatic and non-climatic stressors are inherently interconnected and interact in a complex and complicated manner to enhance social-ecological systems (SES) vulnerability. Inadequate drought adaptation measures further marginalise farmers and ultimately increase their vulnerability to drought risks and loss and damage (L&D). L&D presents itself in various ways, including non-economic L&D and cascading effects that are impacting upon future generations. With projected climate change and the likelihood of impacts worsening, there is an urgency to address L&D more coherently through crucial policy interventions.

Assessing the Effectiveness of Cyber Domain Controls When Conducting Cybersecurity Audits: Insights from Higher Education Institutions in Canada

This study validates a comprehensive cybersecurity audit model through empirical analysis in three higher education institutions in Canada. The research aims to enhance cybersecurity resilience by assessing the effectiveness of cybersecurity controls across diverse educational environments. Given the increasing frequency and sophistication of cyberattacks targeting educational institutions, this research is essential to ensure the protection of sensitive academic and personal data. Data were collected through detailed audits involving system vulnerabilities, compliance with security policies, and incident response management at each institution. The findings underscore the importance of tailored cybersecurity strategies and continuous auditing to mitigate cyber risks in the Canadian higher education sector. This study contributes to the field by validating a versatile audit tool that can be adapted to various institutional contexts, promoting enhanced cybersecurity practices and evaluating the effectiveness of cybersecurity safeguards across the higher education sector in Canada. The results of the audit model validations provide the cybersecurity maturity rating of each institution. Further research is recommended to refine the model and explore its application in other industries and sectors.

CARL: Unsupervised Code-Based Adversarial Attacks for Programming Language Models via Reinforcement Learning

Code based adversarial attacks play a crucial role in revealing vulnerabilities of software system. Recently, pre-trained programming language models (PLMs) have demonstrated remarkable success in various significant software engineering tasks, progressively transforming the paradigm of software development. Despite their impressive capabilities, these powerful models are vulnerable to adversarial attacks. Therefore, it is necessary to carefully investigate the robustness and vulnerabilities of the PLMs by means of adversarial attacks. Adversarial attacks entail imperceptible input modifications that cause target models to make incorrect predictions. Existing approaches for attacking PLMs often employ either identifier renaming or the greedy algorithm, which may yield sub-optimal performance or lead to high inference times. In response to these limitations, we propose CARL, an unsupervised black-box attack model that leverages reinforcement learning to generate imperceptible adversarial examples. Specifically, CARL comprises a programming language encoder and a perturbation prediction layer. In order to achieve more effective and efficient attack, we cast the task as a sequence decision-making process, optimizing through policy gradient with a suite of reward functions. We conduct extensive experiments to validate the effectiveness of CARL on code summarization, code translation, and code refinement tasks, covering various programming languages and PLMs. The experimental results demonstrate that CARL surpasses state-of-the-art code attack models, achieving the highest attack success rate across multiple tasks and PLMs while maintaining high attack efficiency, imperceptibility, consistency, and fluency.

Концепт «Європа» в українському суспільному дискурсі наприкінці першої чверті 21 ст.

The article analyzes the conceptual features of the European phenomenon that formed in Ukrainian public discourse at the end of the first quarter of the 21st century. The purpose of the article is to determine the content of the concept of "Europe" that was in Ukrainian public discourse at the end of the first quarter of the 21st century, when the civilizational choice of Ukrainians received a new impetus for realization through the decision of the European Council to begin negotiations on Ukraine's accession to the European Union. The main research method is conceptual analysis, which involves studying various substrate mental units – representations, scenarios, and gestalts associated with the phenomenon of Europe in the collective consciousness of Ukrainian society. Additionally, the method of discourse analysis is applied at the stage of processing the mass media objectification of the concept of "Europe". Results. The following conceptual features have been identified: Europe as a space for the comprehensive realization of the individual; well-being and freedom as fundamental civilizational principles aspired to by Ukrainian society; democracy built on the rule of law and respect for individuals; partnership relations among all states and nations without domination by the stronger; unity of diverse cultures; civilizational choice for progress into the future; an influencing factor on Ukraine; a state of crisis associated with the weaknesses of liberal democracy on the one hand, and the dominance of consumer interests in European societies on the other; and the vulnerability of the political system to hybrid expansion by totalitarian and autocratic regimes. Conclusion. The analysis of the verbal explication of the conceptual features of the phenomenon of Europe in Ukrainian public discourse leads to the conclusion that the content of the studied linguistic construct in the collective consciousness of Ukrainians at the end of the first quarter of the 21st century is formed by the following main frames: institutional Eurocentrism with the system-forming role of the European Union, civilizational values, socio-cultural unity of diversity, production culture, and Eurointegration aspirations as a path to the future.

Enhancing anomaly detection: A comprehensive approach with MTBO feature selection and TVETBO[sbnd]Optimized Quad-LSTM classification

The rapid growth of computer networks has heightened system vulnerabilities, challenging traditional Machine Learning (ML) techniques in network anomaly detection due to the abundance of data, imbalanced attack classes, and large-scale datasets. This paper introduces a novel approach leveraging the AWID3 dataset, utilizing advanced preprocessing techniques such as data splitting and normalization. We propose the Mountaineering Team-Based Optimization (MTBO) algorithm for robust feature selection, inspired by social behavior and human collaboration. Additionally, we deploy a Quad-LSTM framework, integrating ConvLSTM, CNN-LSTM, E3D-LSTM, and SAConvLSTM models, fine-tuned by the Technical and Vocational Education and Training-Based Optimizer (TVETBO). Our approach achieves a remarkable 99.91 % accuracy, significantly surpassing existing models, thus enhancing network security by effectively addressing the complexities of anomaly detection.

Modeling Interdependent Infrastructure System Vulnerability with Imprecise Information Using Two Fuzzy Inference Systems

Infrastructure systems play important roles in economic development and the social quality of life. Interdependencies exist between infrastructure systems: a functional disruption in one system can affect dependent systems, thereby escalating the impacts. It is vital to properly model interdependencies to understand the full impacts of disruptive events on infrastructure systems. Quantitative data on infrastructure interdependency is often difficult to obtain or unavailable for a variety of reasons. To overcome quantitative data scarcity issues, qualitative subject expert knowledge has been used in interdependency analysis, primarily in the form of linguistic responses. Linguistic data is susceptible to uncertainties arising from variations in intended meanings, which may yield inaccurate results. This paper proposes a framework to address this problem using two fuzzy inference systems to model event-specific, network-wide infrastructure failures. The first fuzzy inference system models the damage induced by interdependencies using verbal descriptions. The second fuzzy inference system accounts for synergistic, compounding effects of multiple incidences of indirect damage caused by interdependencies. A case study is conducted to demonstrate the applicability of the proposed methodology using electric and gas distribution networks in the United Kingdom. Sensitivity analyses are performed to show the flexibility of the fuzzy inference systems. The results show that the proposed method can model the interdependency and vulnerability of infrastructure systems using fuzzy inference systems to handle imprecise input. The proposed framework may assist practitioners in better understanding the interdependency and vulnerability of infrastructure systems, and in making more informed decisions to reduce losses resulting from disruptive events.

Generated or Not Generated (GNG): The Importance of Background in the Detection of Fake Images

Facial biometrics are widely used to reliably and conveniently recognize people in photos, in videos, or from real-time webcam streams. It is therefore of fundamental importance to detect synthetic faces in images in order to reduce the vulnerability of biometrics-based security systems. Furthermore, manipulated images of faces can be intentionally shared on social media to spread fake news related to the targeted individual. This paper shows how fake face recognition models may mainly rely on the information contained in the background when dealing with generated faces, thus reducing their effectiveness. Specifically, a classifier is trained to separate fake images from real ones, using their representation in a latent space. Subsequently, the faces are segmented and the background removed, and the detection procedure is performed again, observing a significant drop in classification accuracy. Finally, an explainability tool (SHAP) is used to highlight the salient areas of the image, showing that the background and face contours crucially influence the classifier decision.

Results of a Survey of UK Farmers on Food System Vulnerability over the Short and Long Term

We report the results of a survey of farmers and landowners to identify the most likely potential food system disruption scenarios for the UK and compare these with a previous expert elicitation with a much wider set of food system stakeholders. We found that 60% of farmers think a Societal Event in which 1 in 2000 people are injured in the UK is at least 20% likely to occur over the coming decade. Over a timeframe of 50 years, this increased to almost 90% of farmers. These results show that farmers and landowners are considerably more concerned about the vulnerability of the food system in the UK than the wider group of food system experts are. Farmers agreed with experts on the majority of potential causes of such vulnerability, which are climate change, trade policies (import and export), competition for land and ecological collapse (over 50 years). However, they also highlighted the importance of the power structure within the food system, with large corporations supplying to, or buying from, farmers creating lower revenue, making farming an unsustainable business. We conclude that an urgent systematic review of potential interventions that would improve resilience be conducted by the UK Government, in partnership with farmers.

Analysis of security vulnerabilities in vehicle On-Board Diagnostic systems

Analysis of security vulnerabilities in vehicle On-Board Diagnostic systems

Neonatal amygdala volumes, procedural pain and the association with social-emotional development in children born very preterm.

Very preterm birth (< 32 weeks' gestational age) is associated with later social and emotional impairments, which may result from enhanced vulnerability of the limbic system during this period of heightened vulnerability. Evidence suggests that early procedural pain may be a key moderator of early brain networks. In a prospective cohort study, neonates born very preterm (< 30 weeks' gestation) underwent MRI scanning at term-equivalent age (TEA) and clinical data were collected (mechanical ventilation, analgesics, sedatives). Procedural pain was operationalized as the number of skin breaking procedures. Amygdala volumes were automatically extracted. The Strengths and Difficulties questionnaire was used to assess social-emotional outcomes at 5 years of age (mean age 67.5 months). General linear models were employed to examine the association between neonatal amygdala volumes and social-emotional outcomes and the timing and amount of procedural pain exposure (early within the first weeks of life to TEA) as a moderator, adjusting for biological sex, gestational age, 5-year assessment age, days of mechanical ventilation and total cerebral volumes. A total of 42 preterm infants participated. Right amygdala volumes at TEA were associated with prosocial behaviour at age 5 (B = -0.010, p = 0.005). Procedural pain was found to moderate the relationship between right amygdala volumes in the neonatal period and conduct problems at 5 years, such that early skin breaking procedures experienced within the first few weeks of life strengthened the association between right amygdala volumes and conduct problems (B = 0.005, p = 0.047). Late skin breaking procedures, experienced near TEA, also strengthened the association between right amygdala volumes and conduct problems (B = 0.004, p = 0.048).

Measles resurgence in Armenia: unmasking health system vulnerabilities and crafting comprehensive eradication strategies - a review.

Measles is an acute febrile illness associated with rashes, fever and life-threatening complications. It is a vaccine-preventable disease with the Measles Mumps Rubella(MMR) vaccine but a recent global trend unveils a resurgence of measles in various parts of the world including Armenia. Measles was declared eliminated from Armenia in 2021 before reports of local outbreaks surfaced in 2023. The WHO identified discrepancies in the vaccination coverage during COVID-19. Measles also poses a great financial burden as a public health issue worldwide. Sociocultural factors impacting measles transmission are maternal education, healthcare access, migration, vaccine hesitancy, and socioeconomic status (Table 3). Efforts to control and eradicate measles from Armenia are run by a collaborative approach of national and international health bodies such as United Nations Children's Fund (UNICEF), WHO, Global Vaccine Alliance (GAVI), and the Ministry of Health. Thus, the resurgence of measles can be managed through widespread patient education, innovative approaches, strengthening the healthcare system and addressing vaccine hesitancy, sociocultural barriers, and humanitarian emergencies. This review investigates the complicated dynamics of measles inside Armenia's health system in depth. A thorough examination of measles incidence and immunization patterns highlights the shift from few cases and high vaccination rates to a comeback caused by imported viruses. The causes of its recurrence have been thoroughly investigated, including reduced immunization programs and vaccine hesitancy. The research delves further into the 'One Health' idea, assessing the possibility of interspecies transmission among nonhuman primates and examining the environmental factors that influence measles transmission. Among the challenges are weaknesses within Armenia's health system as well as the possibility of interruptions from the COVID-19 outbreak. In measles vaccination status, the combination of maternal education, postnatal care, and socioeconomic variables exposes the larger drivers at work. The study concludes with a comprehensive set of public health policy recommendations covering vaccination promotion, surveillance, healthcare provider education, public awareness, international collaboration, data analysis, law enforcement, emergency preparedness, research, and coordination. The research sheds light on the tangled web of measles dynamics, health system resilience, and contextual subtleties via this multifaceted approach, inviting readers to investigate the multiple strategies required for eliminating measles in Armenia.

Vulnerabilities of the neoliberal global food system: The Russia–Ukraine War and COVID‐19

AbstractThe coronavirus 2019 (COVID‐19) pandemic and the Russia–Ukraine war have demonstrated that the neoliberal system is unstable during global crises. In times of crisis, exporter countries adopt protectionist policies in the form of export restrictions to safeguard their local food supply and curb inflation. Consequently, low‐income countries might find themselves unable to access essential food products. In this regard, the aftermath of the COVID‐19 pandemic and the Russia–Ukraine war has gradually increased export restrictions, causing severe food supply disruptions. In particular, import‐dependent countries cannot access essential food products and face famine. To this point, this study explores the vulnerabilities of neoliberalism when exporter countries turn to protectionism. Moreover, it asks whether food sovereignty and self‐sufficiency could act as a safeguard for import‐dependent states against such vulnerabilities. In doing so, the study aims to contribute to the literature by linking protectionism with export restrictions, diverging from the more common association of protectionism with solely import restrictions.

When is lethal deceptive pollination maintained? A population dynamics approach.

Not all plant-pollinator interactions are mutualistic, and in fact, deceptive pollination systems are widespread in nature. The genus Arisaema has a pollination system known as lethal deceptive pollination, in which plants not only attract pollinating insects without providing any rewards, but also trap them until they die. Many Arisaema species are endangered from various disturbances including reduction in forest habitat, modification of the forest understory owing to increasing deer abundance, and plant theft for horticultural cultivation. We aimed to theoretically investigate how lethal deceptive pollination can be maintained from a demographic perspective and how plant and pollinator populations respond to different types of disturbance. We developed and analysed a mathematical model to describe the population dynamics of a deceptive plant species and its victim pollinator. Calibrating the model based on empirical data, we assessed the conditions under which plants and pollinators could coexist, while manipulating relevant key parameters. The model exhibited qualitatively distinct behaviours depending on certain parameters. The plant becomes extinct when it has a low capability for vegetative reproduction and slow transition from male to female, and plant-insect co-extinction occurs especially when the plant is highly attractive to male insects. Increasing deer abundance has both positive and negative effects because of removal of other competitive plants and diminishing pollinators, respectively. Theft for horticultural cultivation can readily threaten plants whether male or female plants are frequently collected. The impact of forest habitat reduction may be limited compared to that of other disturbance types. Our results have emphasised that the demographic vulnerability of lethal deceptive pollination systems would differ qualitatively from that of general mutualistic pollination systems. It is therefore important to consider the demographics of both victim pollinators and deceptive plants to estimate how endangered Arisaema populations respond to various disturbances.

Whole-body exposure to filtered fraction of diesel exhaust induced localized testicular damage through attenuated functional response of glutathione-s-transferase in adult male Wistar rats

The possible vulnerability of the male reproductive system to environmental pollutants such as air pollution necessitates a thorough investigation of the underlying mechanisms involved in the dysregulation of male reproductive function. The present study was designed to investigate the influence of the filtered fraction of diesel exhaust (predominantly comprising gases) on male reproductive function in Wistar rat model. Adult male rats were randomly assigned into three groups (n=8/group): Control (unexposed) group (CG-A), the Clean air group in WBE chamber (CAG-A), and Filtered diesel exhaust group in WBE chamber (FDG-A). The exposure protocol for CAG-A and FDG-A was 6 h/day x 5d/week x 6 weeks,evaluation of sperm parameters, testicular histopathology, quantification of hormones (testosterone, LH, FSH, 17β-Estradiol, and prolactin), and GST levels were performed. Results showed that WBE to FDE leads to a significant decline in sperm concentration (p=0.008, CG-A vs FDG-A; p=0.014, CAG-A vs FDG-A), motility (p=0.008, CG-A vs FDG-A; p=0.029, CAG-A vs FDG-A), serum testosterone (p=0.024, CG-A vs FDG-A; p=0.007, CAG-A vs FDG-A), testicular testosterone (p=0.008, CG-A vs FDG-A; p=0.028, CAG-A vs FDG-A), 17β-Estradiol (p=0.007, CG-A vs FDG-A), and GST levels (p=0.0002, CG-A vs FDG-A; p=0.0019, CAG-A vs FDG-A). These findings demonstrate the disruption of testosterone-estradiol balance in the intratesticular milieu without significant alterations in other principal pituitary hormones in adult rats exposed to FDE. The predominant presence of gaseous components in FDE can cause testicular damage due to oxidative imbalance. This underscores the causality of FDE exposure and impaired male reproductive outcomes.

Adversarial Attack and Defense in Deep Ranking.

Deep Neural Network classifiers are vulnerable to adversarial attacks, where an imperceptible perturbation could result in misclassification. However, the vulnerability of DNN-based image ranking systems remains under-explored. In this paper, we propose two attacks against deep ranking systems, i.e., Candidate Attack and Query Attack, that can raise or lower the rank of chosen candidates by adversarial perturbations. Specifically, the expected ranking order is first represented as a set of inequalities. Then a triplet-like objective function is designed to obtain the optimal perturbation. Conversely, an anti-collapse triplet defense is proposed to improve the ranking model robustness against all proposed attacks, where the model learns to prevent the adversarial attack from pulling the positive and negative samples close to each other. To comprehensively measure the empirical adversarial robustness of a ranking model with our defense, we propose an empirical robustness score, which involves a set of representative attacks against ranking models. Our adversarial ranking attacks and defenses are evaluated on MNIST, Fashion-MNIST, CUB200-2011, CARS196, and Stanford Online Products datasets. Experimental results demonstrate that our attacks can effectively compromise a typical deep ranking system. Nevertheless, our defense can significantly improve the ranking system's robustness and simultaneously mitigate a wide range of attacks.

Artificial intelligence, adversarial attacks, and ocular warfare

PurposeWe explore the potential misuse of artificial intelligence (AI), specifically large language models (LLMs), in generating harmful content related to ocular warfare. By examining the vulnerabilities of AI systems to adversarial attacks, we aim to highlight the urgent need for robust safety measures, enforceable regulation, and proactive ethics. DesignA viewpoint paper discussing the ethical challenges posed by AI, using ophthalmology as a case study. It examines the susceptibility of AI systems to adversarial attacks and the potential for their misuse in creating harmful content. MethodsThe study involved crafting adversarial prompts to test the safeguards of a well-known LLM, OpenAI's ChatGPT-4.0. The focus was on evaluating the model's responses to hypothetical scenarios aimed at causing ocular damage through biological, chemical, and physical means. ResultsThe AI provided detailed responses on using Onchocerca volvulus for mass infection, methanol for optic nerve damage, mustard gas for severe eye injuries, and high-powered lasers for inducing blindness. Despite significant safeguards, the study revealed that with enough effort, it was possible to bypass these constraints and obtain harmful information, underscoring the vulnerabilities in AI systems. ConclusionAI holds the potential for both positive transformative change and malevolent exploitation. The susceptibility of LLMs to adversarial attacks and the possibility of purposefully trained unethical AI systems present significant risks. This paper calls for improved robustness of AI systems, global legal and ethical frameworks, and proactive measures to ensure AI technologies benefit humanity and do not pose threats.