Multi-model large-scale AI framework for avian influenza surveillance and preparedness: Harnessing large language models to enhance risk communication, real-time decision support, and public health response strategies.

Influenza A viruses infect a wide range of animals including poultry, wild birds, pigs, horses, dogs, and marine mammals. Influenza in animals threatens animal health and welfare, agricultural productivity, public health, food security, and the livelihoods of farmers across the globe. The recent H1N1 pandemic of 2009, continuous reporting of zoonotic infections with highly pathogenic avian influenza (HPAI) H5N1 and other avian and swine influenza viruses, such as H3N2v, raise ongoing concerns regarding the emergence of zoonotic viruses with pandemic potential.1 Different strains of influenza A virus show host specificity and are often defined by the species in which they are initially found to be circulating. However, over time the situation becomes more complicated as the viruses continuously evolve, through mutation and reassortment, and in some cases are transmitted from species to species. The next major pandemic is likely to be caused by a strain of influenza virus that is new to that generation; a virus to which the human population has little or no immunity. Almost certainly such a strain would contain genes from influenza viruses that have been circulating in animals. A better understanding of the mechanisms responsible for interspecies transmission, and information on host adaptation and pathogenicity are needed to allow more informed assessment as to when and where the next pandemic may arise. Timely identification of viruses with pandemic potential could ultimately reduce the impact of a new pandemic. With current levels of knowledge and surveillance, it is not possible to accurately assess geographic location of all animal influenza viruses, the production systems in which they circulate, and which of these viruses may be transmitted to and adapt in the human population. This makes predictions about which strains of influenza virus to select for vaccines for pandemic preparedness a challenging task. With mortality rates reaching up to 100% in affected populations, HPAI viruses continue to have a devastating impact in poultry populations. Low-pathogenic avian influenza (LPAI) viruses also have a significant and variable impact on poultry production, depending on the strain and health status of the birds, and may evolve into HPAI viruses, for example, the recent HPAI H7N3 outbreak in Jalisco, Mexico.2, 8 These impacts together with consequential cost of control and trade measures, aimed at preventing further spread, lead to huge economic costs. There is often a greater impact on countries with a lower Gross Domestic Product (GDP), who rely on agriculture for economic development and for sustenance. Early detection of HPAI and H5 and H7 LPAI virus infections in poultry is essential for an effective response which relies on a combination of classic control measures (culling infected flocks and high risk contacts, disinfection, biosecurity, and trade measures) and, where appropriate, vaccination.3 Delays in detection lead to spiraling costs to keep epidemics under control and make the disease more difficult to eradicate. In the case of H5N1 HPAI, effective control in the animal source is needed to reduce the public health risk. Developing countries do not always have the resources to maintain the infrastructure and technical capacities needed for rapid and accurate diagnostic testing and characterization of viral strains. These countries rely on international reference laboratories to test their specimens and characterize the viruses. Research is needed to develop accurate, cheap, and robust diagnostic tests to ensure that the disease is detected early, with sufficient confidence, to allow timely initiation of effective response measures. For initial disease confirmation, both sensitivity and specificity of tests are essential because the implications of false-negative and false-positive results can be considerable. Animal influenza is not only a constraint for agriculture and food production. Equine influenza is an ongoing problem for companion and competition horses and has a huge impact on the horse racing industry. The 2007 equine influenza outbreak in Australia is estimated to have cost the horse racing and gambling industries 3·6 billion Australian dollars in lost revenue.4 Research, accompanied by increased global surveillance, is needed to ensure that the equine sector is able to access effective up-to-date vaccines so that many can continue to benefit from the enjoyment and financial gains that horse sports offer. Some suggest that historic accounts showing a temporal relationship between respiratory disease in horses and humans may implicate influenza viruses.5 However, these accounts date back to a time before influenza viruses had been isolated, and a clear link to influenza as we know it would be difficult to prove. Influenza viruses of subtype H3N8 currently circulating in the horse population have also crossed the species barrier and become established in dogs. However, this subtype and other strains of equine influenza viruses do not appear to be a significant zoonosis despite intense exposure of owners to their horses and dogs, and vice versa. An understanding of the underlying reasons for this may help to explain why other influenza viruses are zoonotic. Effective and cost-effective control require targeting resources for optimal impact. Research is needed to gain an understanding of how control measures can be better targeted. More rapid control in the animal population will limit impacts on animal health and public health when the influenza virus is zoonotic and will minimize costs in terms of production losses and access to international markets. Currently, vaccination does not always prevent infection of birds nor does it prevent infected vaccinated birds from shedding virus. If vaccines are not adequately matched antigenically to circulating field viruses and at least 60–80% of the susceptible populations are immunized, vaccination as a program will not be effective. Further research is needed to improve the effectiveness of the control measures themselves, such as vaccination, and to provide improved access to resources needed for control. Despite unprecedented levels of international investment to support avian influenza surveillance between 2004 and 2009, global surveillance for influenza viruses in animals is woefully inadequate, with too little being undertaken without adequate coordination. Improvements in surveillance are required to provide early warning for effective control and to inform much needed research.6 As well as surveillance in domestic animals (poultry, horses and pigs), surveillance in wildlife is important; it is now evident that wild birds also play a role in the primary introduction of avian influenza in previously disease-free areas.7 Today, we are not able to fully manage the threats and impacts from animal influenza. In one form or another, influenza A viruses are circulating in every country on the planet. Our understanding of the mechanisms responsible for interspecies transmission, adaptation, and pathogenicity is incomplete, and the methods for risk assessment and disease control are rudimentary. Challenges to reduce threats from animal influenzas are considerable and will only be improved through extensive research and innovation. Continued reports of notifiable avian influenza8 and animal influenza associated human infections highlight the need to monitor influenza viruses in all animal species to better understand their role in causing pandemics and severe zoonotic infections, and in reducing agricultural productivity. OFFLU is the World Organisation for Animal Health (OIE) – Food and Agricultural Organization of the United Nations (FAO) global network of expertise on animal influenza, established in 2005 to address the animal and public health threats from H5N1 HPAI. Since then, its mandate has been extended to cover all animal influenza viruses. OFFLU is unique in that its participation comprises a global representation of leading experts in animal influenza including researchers, diagnosticians, policy makers, economists, and epidemiologists. One of OFFLU's core objectives is to advocate for more research, to highlight specific influenza research objectives, promote their development, and to ensure coordination. OFFLU works closely with WHO on all influenza issues at the human–animal interface, including identifying commonly agreed research priorities. Following the OFFLU annual technical meeting in 2010 attended by avian, swine, equine, and public health experts, it was decided that there was an urgent need to develop a Research Agenda to highlight and coordinate research priorities for the animal influenza sector. The Research Agenda highlighted needs in different animal species and at the human–animal interface. It is designed to help policy makers, researchers, and donors ensure that their efforts and resources are targeted to areas where there is an identified need. The agenda should also be used to leverage funds for animal influenza research. In today's world where there is a huge volume of information of variable quality, the OFFLU Research Agenda has been designed to be concise and digestible; it comprises only 11 pages. The animal and human influenza networks share the common goal of reducing public health threats from animal influenza viruses. OFFLU has been working closely with its parent organizations, the OIE and the FAO, and its partner the World Health Organization (WHO) to ensure that its efforts are complementary and well coordinated. In 2008 and 2010, the OIE-WHO-FAO tripartite held joint technical consultations on avian influenza at the human–animal interface in Verona, Italy and discussed other technical interface topics of common interest. The experts identified that more research is needed on modes of transmission, behaviors associated with increased risks of transmission, virologic and ecologic aspects, and viral persistence in the environment to address the human exposure risks to H5N1 infection.9 Importantly in practical terms, OFFLU contributes animal influenza data to the biannual WHO influenza vaccine composition meetings. This information is critical in allowing selection of the most appropriate strains of virus for vaccines to protect against potential zoonotic pandemic influenza, including H5N1 and H9N2 avian influenza.10 Where zoonotic strains of influenza are undergoing antigenic drift in animal populations, the situation is being monitored in real time to allow selection of relevant influenza virus seed strains and antigens for vaccines for public health preparedness. OFFLU and WHO experts are working together to better understand which animal influenza viruses may pose a risk to human health. This ongoing risk assessment is supported by OFFLU's drive to improve and better collate data from avian, swine, and equine influenza surveillance programs world-wide. Animal influenza research is suffering from donor fatigue, and it is a continuing challenge to ensure that sufficient resources can be secured to address the priorities that have been identified, ultimately to improve health and economies. It is considered possible to prevent a human influenza pandemic by identifying influenza viruses with pandemic potential in animal species; this will only be achieved through further influenza research studies in animals. A large body of animal influenza data has been generated in recent years, presenting a real opportunity to increase our understanding of how to identify risk and better control the adverse effects of influenza in animals and at the human–animal interface. Many questions still remain to be answered. Structured and coordinated research toward prioritized goals and objectives will greatly facilitate this. The 'OFFLU Research Agenda' is a first for the animal health sector and will help to steer animal influenza research toward the identified objectives, providing maximum benefits for public and animal health. The influenza research priorities focus on several topics including control and education, diagnostics, epidemiology, immunology and immune responses, pathogenesis, transmission, vaccines and vaccination and virus characteristics and evolution in poultry, wild birds, swine and equine. The full OFFLU research agenda can be viewed at http://www.offlu.net/fileadmin/home/en/publications/pdf/OFFLU_Research_Priorities_photo.pdf. The authors would like to thank David Swayne, Ian Brown, Kristien Van Reeth and Ann Cullinane. The authors have no potential conflicts of interest to declare.

H1N1 after action review: learning from the unexpected, the success and the fear

Building a Strong Foundation for Public Health Transformation.

This article examines the integration of Large Language Models (LLMs) and Generative AI within Enterprise Resource Planning (ERP) systems, highlighting the transformative impact on cognitive process automation. As organizations transition toward autonomous operations, generative AI capabilities embedded within cloud infrastructure and applications create new paradigms for intelligent automation across finance, procurement, and human resources functions. The architectural framework combines foundation models with domain-specific enterprise data to enable conversational interfaces, document understanding, text generation, and real-time decision support. These capabilities extend traditional ERP functionalities beyond conventional automation, creating systems that can reason, learn, and adapt to changing business conditions. The article explores strategic benefits, including cognitive task automation and personalized user assistance, while addressing governance considerations essential for responsible enterprise deployment, such as data sovereignty, auditability frameworks, and tenant isolation mechanisms that balance innovation with appropriate risk management.

Avian influenza A H5N1 continues to be a major threat to global public health as it is a likely candidate for the next influenza pandemic. To protect public health and avert potential disruption to the economy, the Hong Kong Special Administrative Region Government has committed substantial effort in preparedness for avian and pandemic influenza. Public health infrastructures for emerging infectious diseases have been developed to enhance command, control and coordination of emergency response. Strategies against avian and pandemic influenza are formulated to reduce opportunities for human infection, detect pandemic influenza timely, and enhance emergency preparedness and response capacity. Key components of the pandemic response include strengthening disease surveillance systems, updating legislation on infectious disease prevention and control, enhancing traveller health measures, building surge capacity, maintaining adequate pharmaceutical stockpiles, and ensuring business continuity during crisis. Challenges from avian and pandemic influenza are not to be underestimated. Implementing quarantine and social distancing measures to contain or mitigate the spread of pandemic influenza is problematic in a highly urbanised city like Hong Kong as they involved complex operational and ethical issues. Sustaining effective risk communication campaigns during interpandemic times is another challenge. Being a member of the global village, Hong Kong is committed to contributing its share of efforts and collaborating with health authorities internationally in combating our common public health enemy.

Shifting Public Health Priorities and the Global Effort to Prevent a Bird Flu Pandemic

e13623 Background: Bone cancer is a complex and challenging disease to diagnose and treat in clinical practice. Recently, generative AI, especially large language models (LLMs), has demonstrated potential as a decision support tool for cancer. However, most implementations have overlooked the integration of available cancer guidelines, such as the NCCN Bone Cancer Guidelines, in fine-tuning the outputs of generative AI models. Incorporating these guidelines into LLMs presents an opportunity to harness the extensive clinical knowledge they contain and improve the decision-support capabilities of the model. Methods: In this study, the aim is to enhance the LLM with cancer clinical guidelines to enable accurate medical decisions and personalized treatment recommendations. Therefore, we introduce a novel method for incorporating the NCCN Bone Cancer Guidelines into LLMs using a Binary Decision Tree (BDT) approach. The approach involves constructing a BDT based on NCCN Bone Cancer Guidelines, where internal nodes represent decision points from the Guidelines, and leaf node signify final treatment suggestions. Then the LLM makes decision at each internal node, considering a given patient's characteristics, and guides toward a treatment recommendation in the leaf node. To assess the efficacy of Guideline-enhanced LLMs, an oncologist from our team created 11 hypothetical osteosarcoma patients’ medical progress notes. Each note contains their demographics, medical history, current illness, physical exams, diagnostic tests. We tested three LLMs in the implementation (GPT-4, GPT-3.5, and PaLM 2) and compared the LLM-generated treatment recommendations with the gold standard treatment across four runs with different random seeds (random seeds is a setting to control the LLM outputs). The results are reported as the average of four runs. The original LLMs are used as baseline methods for comparison. Results: The table below provides a comparison between the performance of original LLMs and those augmented with cancer guidelines for osteosarcoma treatment recommendations. We can observe that the PaLM 2 model demonstrated superior performance compared to its counterparts, underscoring the effectiveness of integrating cancer guidelines into LLMs for decision support. Conclusions: The clinical decision support capabilities of the LLMs are promising when enhanced by NCCN Bone Cancer Guidelines using our approach. To fully exhibit the potential of our proposed method as a clinical decision support tool, further investigation into other subtypes of bone cancer should be conducted in the future study. [Table: see text]

Lessons from the COVID-19 Pandemic—Unique Opportunities for Unifying, Revamping and Reshaping Epidemic Preparedness of Europe’s Public Health Systems

Nations set out a global plan for influenza action

This study explores specific insights of university professors using large language models (LLMs) across their different roles.Using think-aloud protocols, we interviewed ten Norwegian professors from mathematics and natural sciences actively engaged in a project integrating these tools in teaching.The study surfaces emergent applications of LLMs in higher education with professors' reflections on using LLMs in teaching & research practices. InntoductionAs university leaders advocate integrating generative AI tools for teaching and learning, professors are envisioning how they might augment their scholarship.These explorations are learning transfer (Pea, 1987) as they integrate LLMs and generative AI in their research.When real-world data are limited or hard to obtain, researchers are using generative AI to generate synthetic data (Figueira & Vaz, 2022), to augment data analysis and predictions, e.g., synthetic human data posture to overcome data scarcity with recruitment/ethics challenges (Dindorf et al., 2024).Augmentation of data annotation through a human-AI collaboration approach in computational social science research is an emergent LLM application (Ziems et al., 2024).In computer science, LLMs are experienced as revolutionary advances for research and teaching, enabling fast writing of high-quality code (Zelikman et al., 2022).Researchers find LLMs can support academic work-as writing/editing tools benefiting researchers from all backgrounds, particularly non-native English speakers; assisting in grant writing; in reviewing papers/grant proposals; programming aids; facilitating course design, lesson plans, and assessments (Meyer et al., 2023).Recognizing professors typically engage in three main tasks-scholarship, teaching, and service/leadership-we undertook open-ended faculty explorations at a Higher Education Institution where professors typically assume all three roles and now seek to integrate LLMs into their teaching.What specific insights do university professors derive from using LLMs across their different professorial roles? Method Research ParticipantsAfter ethical approval from the university's ethics committee, 10 of 18 Norwegian university professors from mathematics and natural sciences involved in brainstorming ways to integrate LLMs in teaching joined the study.The first author engaged as a collaborative researcher.At project launch (Aug'23), most professors had tried at least one LLM-powered tool (ChatGPT, Bing, Bard); 3 had not tried any.Acknowledging LLMs and tools they power are not the same, we use their names interchangeably.

First Line of Defense

In early 2020, public health workers across the United States were called to respond to an emerging threat: COVID-19. Seemingly overnight, COVID-19 became a pandemic and, as many transitioned from offices and schools to home-based settings, essential workers braved the risk of infection to face the emergency and maintain essential health services. The pace of the response and the scale of the loss of life in the United States were unprecedented in recent history. Public health workers demonstrated their dedication to their mission by rising to these challenges, often stretching themselves beyond their capacity to meet the demands of the crisis. In the months and years since the first case of COVID-19, there has been a seismic shift in the way society engages the public health workforce. The 2021 Public Health Workforce Interests and Needs Survey (PH WINS) provides a snapshot of the burden carried by the public health community. Even before the emergence of COVID-19, many public health workers have moved from one emergency response to the next with little pause for recovery, exacerbating systemic challenges. Our experience managing multiple simultaneous and overlapping public health emergencies has demonstrated the fragility of our public health infrastructure and eroded public trust. As we face an increasing frequency and severity of public health disasters in the contexts of climate change and organized health disinformation, a deepening distrust of science has forever changed the nature of our work. In addition, as the field of public health recognizes systemic racism as a public health crisis and takes meaningful steps to dismantle it, it is threatened with mounting hostility from outside and within our government structures. These experiences, reflected in PH WINS 2021, have also brought to light a world of opportunities to build a better public health system that will sustain through and beyond the emergencies of the future (Table). TABLE - Pathways to Resilience Build internal tracks to leadership for staff from communities that are heavily impacted in emergencies. Transform COVID-19 temporary public health workers into a new public health workforce. Develop a dynamic public health emergency response infrastructure. Build resilience in essential basic public health functions. Create trauma-responsive environments for the public health workforce. Build Internal Tracks to Leadership for Staff From Communities That Are Heavily Impacted in Emergencies The COVID-19 pandemic highlighted long-standing health inequities. Structural oppression creates community- and neighborhood-level health vulnerabilities before, during, and after public health emergencies1–3; however, the makeup of our current public health leadership is limited by generations of exclusion of people from the communities that could most benefit from public health programming. Historic definitions of expertise in public health exclude some of the most critical “qualifications”—those gained by lived experience. During the COVID-19 response, health departments in need of critical community-level information often did not have to look far; highly adept individuals from heavily impacted communities were already part of the public health workforce, but their indispensable skills and knowledge were not measured in their job titles or work assignments. Emergency responses can exacerbate or dismantle long-standing inequities perpetuated by systemic racism within governmental agencies. Public health must embrace the opportunity to unravel systems of oppression by identifying staff who live in the hardest-hit communities, uplifting them into leadership roles, and developing intentional partnerships with communities. Such approaches during emergency response can give staff the opportunity to gain leadership experience, develop skills in rapid participatory action research and qualitative methods, and forge the robust partnerships with communities necessary to develop community-relevant solutions and meaningfully improve health equity.3 Transform Temporary Public Health Workers Who Worked in COVID-19 Into a New Public Health Workforce Public health emergencies create critical opportunities to invest in a dangerously underfunded system.4,5 Funding streams emerging after public health emergency responses should be invested in remediating the systemic challenges that prolong and exacerbate emergencies, particularly by building a workforce from impacted communities. During COVID-19, while rapid scaling of a temporary workforce resulted in public health gains, many workers hired through an influx of emergency funding were laid off as the response deescalated. For example, NYC Trace, the largest contact tracing operation in the country, rapidly hired thousands of New Yorkers at a time when unemployment was high and relatively safe remote jobs were scarce. Selective recruitment from highly impacted communities at that time increased the acceptability of services while both investing in impacted communities and cultivating a pool of trained public health workers.3 Unless these workers are reintegrated into the public health workforce, the field risks losing their newly gained expertise. Given massive workforce shortages experienced by public health departments across the country, decision makers could look to approaches adopted by global humanitarian response programs and the Public Health Corps and consider developing national assignments to fill short- and long-term gaps. Not only would this fulfill urgent, mounting workforce needs and support the development of the public health careers for many essential pandemic workers but this workforce would also bring greater community-centered knowledge and practice to the field, nationwide. Develop a Dynamic Public Health Emergency Response Infrastructure As we emerge from the acute phases of the COVID-19 pandemic, we have an opportunity to revisit the structure of our public health emergency response systems, building upon a strong backbone of essential public health competencies. Across the country, public health workers were reassigned from their day jobs to emergency roles, often for years on end. PH WINS 2021 shows that all public health program areas contributed at least 20% of their workforce time to COVID-19 response efforts.6 This approach creates gaps in ongoing public health activities, burnout among employees, and limited institutional memory to inform future emergencies. To be responsive, emergency response systems need long-term investments to modernize and improve core public health functions. There is a clear need to institutionalize lessons learned and quickly integrate corrective actions. During and after a response, innovation and accountability in “hotwashes” and after-action reports could support this process but only if evaluation is followed by swift, measurable action. The resultant dynamic response system would support the translation of knowledge and skills from one response to another and produce a more efficient allocation of public health resources toward foundational infrastructure required in emergencies and core public health services. Build Resilience in Essential Basic Public Health Functions COVID-19's devastating legacy exceeds its direct morbidity and mortality. As case counts exploded in early 2020, routine public health operations and health care services screeched to a halt. While organizations worked around the clock to create electronic versions of services previously provided in person and emergency response programs incorporated wraparound services, not all services translated effectively and not all people had the same access to digital platforms. This has resulted in unprecedented setbacks in disease prevention, social-emotional learning, education, chronic and acute disease management, life expectancy, and more. At the same time, we are not seeing any reprieve from public health emergencies. These 2 are not separate: widening gaps in our public health systems multiply the impacts of public health disasters when they hit.1–3 Public health agencies must build capacity in core public health functions including surveillance and accompanying data informatics, data and risk communication systems, and robust mental health infrastructure. Instead of skirting along the edge of staff capacity and shying away from innovative modernization of public health informatics, we must invest in core functions that will not only ensure continuity of basic public health services for all people in all places in the United States but also facilitate smooth pivots to emergency response activities that rely on these same competencies. Create Trauma-Responsive Environments for the Public Health Workforce None of these initiatives or investments can move forward if no one is around to make them happen. PH WINS 2021 shows the devastating toll the past few years have had on the public health workforce, with nearly one-third of the governmental public health workforce considering leaving their organization in the next year for retirement or other reasons.7 Sector-wide burnout and related behavioral health impacts must be addressed while building sufficient capacity to prevent this scale of trauma from recurring in the next disaster. However, trauma is complex, long-lasting, and impacted by culture and environment; mitigation requires awareness and understanding. Now is the moment to build trauma-responsive environments for the public health community. Trauma-responsive environments promote healing and wellness, allow public health workers to access services and space to process their experiences, and support workers to heal. They must be built on a foundation of accessible mental health supports and integrative care approaches, which will be critical to sustaining services long after pandemic recovery funding is exhausted. Behavioral health services can only be accessible if social stigma and other barriers are reduced across the field of public health nationally. Some of these barriers are explicitly embedded in licensing and board requirements, leave policies, and organizational culture.8 Similarly, just as systemic and structurally intersectional inequities persist, so too do historical and intergenerational trauma via oppression, racism, and prejudice. Meaningful action is needed to revise policies that prevent public health workers from accessing behavioral health services. We must do more than ask our workers to “power through” because the next disaster is looming. It is time to improve our systems and way of work to build a foundation for trauma-responsive care throughout disaster response and recovery.

Avian influenza (AI) has had great impacts on global health and socioeconomics in the past two decades. Among many newly emerged influenza viruses, avian influenza viruses (AIVs) have played either indirect or direct role in human infections. The continuous evolvement of AIVs through antigenic drift and genetic reassortment has led to emerging diversities of local virus strains or variants. These dynamic changes of AIVs, even low pathogenic avian influenza (LPAI) viruses, have not only affected poultry health but also resulted in severe and fatal human cases. Moreover, the persistence of these viruses at a population level may lead to outbreaks of AI occurring year after year in a vicious cycle, if they are not completely eradicated. In light of the emergence of novel H7N9 and other subtypes of AIVs in China in recent years, residents of Taiwan who live closely to this neighboring country must be well prepared in case these viruses are imported. Furthermore, the potential for interspecies transmission to humans from either of the two subtypes of avian H5N2 and H6N1 viruses increases the risk assessment of potential public health threats coming from continuous viral mutations. These two AIV subtypes are relevant in chickens and have been endemic in Taiwan for many years, and the newly emerged novel three subtypes (H5N2, H5N3, and H5N8 clade 2.3.4.4) of AIVs that have spread island-wide since January of 2015 must be expeditiously assessed for public health risks. In summary, this article discusses the history, virology, and epidemiology of avian influenza in humans. We believe that fully understanding virological and epidemiological characteristics of avian influenza viruses and the history of past pandemics or major epidemics will help strengthen the effectiveness of prevention and control measures.

Generative AI is reshaping the enterprise technology landscape, offering intelligent automation, insight generation, and contextual understanding capabilities that redefine how businesses handle data. Enterprise data management (EDM) - once constrained by rigid architectures, manual processing, and fragmented governance - can now evolve into a dynamic, self-improving ecosystem through the integration of generative AI. With organizations generating petabytes of data from operations, customer interactions, supply chains, and IoT devices, the need for scalable and intelligent data handling systems has never been greater. Generative AI models, including large language models (LLMs) and multimodal transformers, provide new tools for data ingestion, cleansing, integration, transformation, synthesis, and summarization. By applying generative AI to enterprise data workflows, companies can enhance metadata enrichment, automate data cataloging, improve data lineage tracking, and simplify data governance. These capabilities increase data discoverability, trust, and compliance—core principles of modern data management. Additionally, generative AI supports natural language querying, automates report writing, and generates synthetic data for training and simulation, boosting data availability and operational speed. While generative AI brings immense promise, it also raises concerns around hallucination, model transparency, data privacy, and regulatory compliance. Ensuring responsible AI adoption requires rigorous validation, bias mitigation, and alignment with existing data governance policies. Nonetheless, enterprises that embrace generative AI can unlock superior decision-making, improve productivity, and democratize data access across technical and non-technical users. This white paper explores the opportunities, challenges, architectural considerations, and best practices for embedding generative AI into enterprise data management. Through industry examples and forward- looking analysis, it offers a roadmap for transforming data operations and maximizing enterprise intelligence in the era of AI. Keywords: Generative AI, Enterprise Data Management, LLMs, Data Governance, Metadata, Data Cataloging, Synthetic Data, Data Lineage, Natural Language Processing, Responsible AI

BACKGROUND During the COVID-19 pandemic, the rapid spread of misinformation on social media created significant public health challenges. Large language models (LLMs), pretrained on extensive textual data, have shown potential in detecting misinformation, but their performance can be influenced by factors such as prompt engineering (ie, modifying LLM requests to assess changes in output). One form of prompt engineering is role-playing, where, upon request, OpenAI’s ChatGPT imitates specific social roles or identities. This research examines how ChatGPT’s accuracy in detecting COVID-19–related misinformation is affected when it is assigned social identities in the request prompt. Understanding how LLMs respond to different identity cues can inform messaging campaigns, ensuring effective use in public health communications. OBJECTIVE This study investigates the impact of role-playing prompts on ChatGPT’s accuracy in detecting misinformation. This study also assesses differences in performance when misinformation is explicitly stated versus implied, based on contextual knowledge, and examines the reasoning given by ChatGPT for classification decisions. METHODS Overall, 36 real-world tweets about COVID-19 collected in September 2021 were categorized into misinformation, sentiment (opinions aligned vs unaligned with public health guidelines), corrections, and neutral reporting. ChatGPT was tested with prompts incorporating different combinations of multiple social identities (ie, political beliefs, education levels, locality, religiosity, and personality traits), resulting in 51,840 runs. Two control conditions were used to compare results: prompts with no identities and those including only political identity. RESULTS The findings reveal that including social identities in prompts reduces average detection accuracy, with a notable drop from 68.1% (SD 41.2%; no identities) to 29.3% (SD 31.6%; all identities included). Prompts with only political identity resulted in the lowest accuracy (19.2%, SD 29.2%). ChatGPT was also able to distinguish between sentiments expressing opinions not aligned with public health guidelines from misinformation making declarative statements. There were no consistent differences in performance between explicit and implicit misinformation requiring contextual knowledge. While the findings show that the inclusion of identities decreased detection accuracy, it remains uncertain whether ChatGPT adopts views aligned with social identities: when assigned a conservative identity, ChatGPT identified misinformation with nearly the same accuracy as it did when assigned a liberal identity. While political identity was mentioned most frequently in ChatGPT’s explanations for its classification decisions, the rationales for classifications were inconsistent across study conditions, and contradictory explanations were provided in some instances. CONCLUSIONS These results indicate that ChatGPT’s ability to classify misinformation is negatively impacted when role-playing social identities, highlighting the complexity of integrating human biases and perspectives in LLMs. This points to the need for human oversight in the use of LLMs for misinformation detection. Further research is needed to understand how LLMs weigh social identities in prompt-based tasks and explore their application in different cultural contexts.