Avian influenza viruses, including HPAI viruses, are characterized by their ability to mutate rapidly, spreading and enabling them to adapt to new hosts. The primary zoonotic transmission routes include direct contact with infected birds and animals, exposure to contaminated environments, like faeces, etc., and the preparation of, and consumption of undercooked poultry (meat or other animal products). Further human activities, such as intensive farming and wildlife trade, amplify the risk of spillover events, events (by increasing the instances of human-animal interactions).

Avian influenza presents a critical global health threat, as indicated in Table 1. According to WHO, from January 1 2003 to December 12, 2024, 954 confirmed human cases of HPAI influenza A (H5N1) virus infection were reported across 24 countries, with 464 fatalities. Cases and fatalities related to other HPAI viruses are indicated in Table 1. While the global trend has been a decline in H5N1 human cases since 2015, recent reports (13, 14) indicate an increasing number of new human infections. For instance, as of January 6, 2025, the United States had reported 66 confirmed human cases of H5N1 since 2024, with one fatality. Historically, cattle were not considered natural hosts for H5N1, however recent cases (such as in the US in 2024) indicate that the virus can infect dairy cows, particularly spreading their mammary glands. In addition, in January 2025, a human case of H5N1 was detected in England, the second symptomatic case in the UK.

It is important to note that although human cases have been relatively low, H5N1 has been spreading extensively among bird populations and has recently been increasingly detected in several mammalian species, including dairy cows and wild animals. For example, in the United States, since 2022, the United States Department of Agriculture Animal and Plant Health Inspection Service has reported HPAI A (H5N1) virus detections in more than 200 mammals (15).

Emerging evidence suggests that certain HPAI strains, particularly from clade 2.3.4.4b, have acquired mutations that may enhance their capacity to infect mammals. Among these, changes in the PB2 gene—such as E627K and D701N—have been identified as key molecular markers that improve viral replication efficiency in mammalian hosts. Although these mutations have been detected in sporadic cases, particularly in mammals exposed to infected birds, there is no conclusive evidence of sustained mammal-to-mammal transmission to date. Nonetheless, the detection of such adaptations underscores the need for heightened genomic surveillance at the human–animal interface, where the potential for zoonotic spillover is greatest.

Another critical genetic determinant in the adaptation of avian influenza viruses to mammalian hosts is the PB2 gene, which encodes one of the three polymerase subunits essential for viral replication. Specific mutations in PB2, notably E627K and D701N, have been repeatedly associated with enhanced viral replication efficiency at the lower temperatures found in the mammalian respiratory tract (16). These mutations can significantly increase the virulence and transmissibility of HPAI viruses in mammals, including humans. Studies following experimental infections and epidemiological investigations of zoonotic cases have consistently highlighted PB2 mutations as key markers for assessing pandemic potential. The capacity of these mutations to facilitate cross-species transmission underlines the importance of incorporating PB2 surveillance into global risk assessment frameworks for avian influenza viruses.

The transition of H7N9 from low pathogenic avian influenza (LPAI) to highly pathogenic avian influenza (HPAI) is marked by the acquisition of a polybasic cleavage site in the hemagglutinin (HA) protein, leading to increased virulence in poultry (17). However, this change does not necessarily enhance transmissibility or severity in humans. Notably, both LPAI and HPAI H7N9 strains have been linked to severe human infections, with case fatality rates around 40% (18).

In Figure 1, we compare the main high-risk HPAI subtypes considering zoonotic risks and mortality rates. The classification of zoonotic risk for HPAI viruses is based on factors such as the frequency and severity of human infections, potential for human-to-human transmission, and genetic characteristics of the virus. While there is not a universally standardized 1-to-5 scale, various health organizations assess these risks to inform public health responses.

The scores calculated in Figure 1 are scientifically based. We supported our classification of zoonotic risks based on two sources (26, 27). We supported our classification of zoonotic risks based on two sources, The Public Health Agency of Canada and WHO. The first one classifies certain influenza A virus subtypes, including H5N1, H5N6, and H7N9, as Risk Group 3 Human Pathogens due to their significant potential to cause serious human or animal disease (27). Similarly, WHO conducts risk assessments for specific HPAI and LPAI strains. In a 2022 assessment, the WHO evaluated the zoonotic risk of the A(H5N1) clade 2.3.4.4b viruses (28–30), considering factors like human cases, virus spread in wild and domestic animals, and genetic mutations. In Figure 1 we provide an indication of these zoonotic risks according to mortality rates, considering the possibility of overestimate of mortality rate due to underestimation of the number of infections especially.

These assessments, while not always presented on a numerical scale, provide a framework for scoring the relative zoonotic risks of different HPAI subtypes.

Innovative approaches to vaccine development are urgently needed and key points include addressing technological development and production challenges to overcome technological gaps. Current egg-based vaccine production is slow and difficult to scale in an emergency, especially during an HPAI virus outbreak that could lead to massive animal culling. While mRNA technology shows promise, questions remain about the duration, breadth of protection and scalability (31). Due to these constraints, our previous study (7) provided evidence indicating that vaccine development for HPAI is lagging, with very few active patents and limited advancements in universal vaccines. This scenario hampers the global capacity to respond effectively to a potential pandemic. Substantial investment in universal influenza vaccines is crucial to address the limitations of current technologies.

In addition to the technological barriers previously discussed, a critical challenge in HPAI vaccine preparedness is the antigenic mismatch between stockpiled vaccines and the strain that causes a future outbreak. This scenario is increasingly likely given the accelerated antigenic drift and shift observed in H5Nx viruses. For example, while clade 2.3.4.4b viruses currently dominate outbreaks in Europe and North America, regions of Asia are still reporting circulation of other clades such as 2.3.2 and 2.3.3. This geographic heterogeneity in dominant clades increases the risk that existing vaccine stockpiles, often produced against earlier strains, will have reduced efficacy if deployed during a novel outbreak elsewhere. Moreover, regulatory and manufacturing timelines for updating avian influenza vaccines lag behind the pace of viral evolution, further compounding this gap. This underscores the urgency for next-generation HPAI vaccines that offer broader cross-clade protection, such as those targeting conserved viral epitopes or using novel platforms like mRNA and recombinant technologies (32, 33).

A recent breakthrough study by researchers from the University of Pittsburgh, in collaboration with the NIH Vaccine Research Center, demonstrated that monkeys pretreated with a moderate dose of the broadly neutralizing antibody MEDI8852¹ were universally protected against HPAI viruses. In addition to confirming the antibody’s efficacy in preventing serious adverse health outcomes, the scientists established the minimum serum concentration required for protection in primate models (34) for further development of universal HPAI vaccines. For instance, deep learning approaches have been used to predict antigenic drift in H5N1 hemagglutinin variants, helping researchers anticipate viral escape mutations (35). Similarly, support vector machines and random forest algorithms have been applied to forecast epitope binding affinity and immunogenicity, enabling more targeted vaccine design (36). Recent studies have also demonstrated the use of neural network-based models to optimize mRNA vaccine sequences for enhanced expression and immunogenicity, which could be adapted for influenza vaccines in the future (37).

Artificial Intelligence (AI) has begun to reshape the landscape of pandemic preparedness, particularly by enhancing the speed and precision of epidemiological surveillance and knowledge governance. In vaccine development, machine learning models can predict antigenic properties, simulate immune responses, and optimize candidate selection, accelerating preclinical pipelines (38, 39). In parallel, AI-powered epidemic intelligence systems can process diverse unstructured data—from genomic databases to wildlife surveillance—to detect abnormal patterns and emerging threats before they affect human populations.

These advances are especially relevant for HPAI, where rapid viral evolution and zoonotic spillover require integrated early warning systems. AI algorithms can assimilate real-time inputs from genomic sequencing, environmental monitoring, animal health records, and social behavior to anticipate outbreak hotspots. This predictive capacity enables health authorities to issue early alerts and prioritize surveillance resources more effectively.

Artificial Intelligence (AI) is emerging as a revolutionary tool, reshaping public health with advancements in analyzing and preventing future pandemic scenarios. Additionally, AI supports robust genomic and antigenic surveillance. Genomic analysis allows tracking of mutations and reassortments that enhance virulence or transmissibility, while antigenic assays assess how well existing immune responses recognize evolving HPAI strains. Integrating both approaches is critical to guide vaccine updates and antiviral strategies.

Beyond surveillance, AI offers transformative applications across operational domains of pandemic preparedness. When adequately designed and integrated within resilient health systems, AI can significantly enhance outbreak forecasting, optimize allocation of medical resources, and support rapid diagnostics. For example, deep learning models have been deployed to predict regional outbreak hotspots based on climatic and migratory bird data, as demonstrated during H5N1 outbreaks in Southeast Asia (40). AI-driven decision support systems have also been used to optimize stockpiling and distribution of antiviral medications and personal protective equipment in real-time emergency settings (41). Additionally, AI-powered diagnostic tools using image recognition and molecular data processing have accelerated point-of-care detection of avian influenza strains in field conditions (42). These efforts must also be grounded in the One Health framework, requiring international coordination and data sharing across human, animal, and environmental health domains to ensure comprehensive risk assessment (7, 26).

Artificial Intelligence (AI) is emerging as a revolutionary tool in public health, with its potential to analyze vast amounts of data, identify trends, and enable informed decision-making. For HPAI, AI could help identify conserved viral epitopes across multiple subtypes, guiding the rational design of universal or broadly protective vaccines. Recent studies have used machine learning models to predict conserved B-cell and T-cell epitopes in H5N1 and H7N9 hemagglutinin and neuraminidase proteins, accelerating preclinical evaluation of cross-protective candidates (35, 43). In pharmacological pipelines, AI-based virtual screening platforms have been applied to search large chemical libraries for molecules with predicted binding affinity to influenza polymerase and neuraminidase targets, significantly reducing lead compound identification time (6). Notably, deep learning frameworks have also been used to repurpose existing antiviral drugs against emerging HPAI strains by predicting off-target antiviral activities (44).

One of the most critical roles of AI is accelerating vaccine and drug discovery. For HPAI, AI can help identify conserved viral epitopes across subtypes, guiding development of universal or broadly protective vaccines. In pharmacological pipelines, it can screen large chemical libraries to identify potential antiviral compounds, reducing the time from discovery to clinical testing. Inclusive governance, characterized by equitable decision-making and transparent data sharing across countries and regions, is fundamental to effective global pandemic preparedness. This includes open access to viral genomic sequences, real-time epidemiological reporting, and collaborative use of AI-driven surveillance platforms to ensure timely detection and response to HPAI threats (45).

AI also facilitates strategic decision-making in resource-constrained scenarios. Algorithms can integrate epidemiological data, health system capacity, and demographic variables to prioritize vaccine allocation, deploy health workers, and anticipate regional surges in infection. Logistics systems enhanced by AI can ensure timely distribution of critical supplies such as PPE, antivirals, and ventilators, especially in underserved areas.

Moreover, AI can support integration of epizootic surveillance with immunization efforts. By linking real-time data from wildlife and livestock with mutation tracking, AI enables targeted containment and adaptive vaccination strategies. This is crucial to prevent the emergence of vaccine-resistant strains or hidden transmission pathways.

However, the success of AI depends on high-quality data inputs, inclusive governance, and ethical frameworks. It must be implemented as part of a broader transdisciplinary preparedness strategy, not as a stand-alone solution. AI’s greatest value lies in its ability to support rapid, data-driven action within a collaborative, globally coordinated response.

Beyond early warning and surveillance, AI also provides valuable tools for accelerating vaccine and drug development, optimizing resource-allocation, and integrating epizootic surveillance systems.

For HPAI, AI could help identify structurally conserved regions of viral proteins across multiple strains, supporting the development of vaccines that offer broad protection. In drug discovery, AI can analyze vast libraries of chemical compounds to identify potential antiviral candidates, drastically reducing the timeline from research to deployment.

In a pandemic scenario marked by high lethality and scarce resources, AI supported by quality data could assist policymakers in making data-driven decisions about resource distribution. AI models could help a timely response to a broad range of indicators, such as population density, healthcare infrastructure, and disease transmission patterns to prioritize vaccine allocation, to deploy healthcare personnel, and to optimize hospital capacities.

AI-based logistics systems can predict areas likely to experience surges in cases, enabling timely delivery of critical supplies like personal protective equipment and ventilators. This proactive approach could ensure that even resource-limited regions are adequately supported.

AI can play a critical role in supporting an integrated “big data” monitoring system that combines epizootic surveillance and immunization, if vaccines are not available. This integrated system can be a powerful tool to prevent and contain outbreaks, identifying viral circulation, monitoring mutations, and detecting early infections in domestic and wild birds. This information is essential for guiding targeted immunization programs and adjusting vaccine formulations to match emerging strains. Molecular diagnostics and genomic sequencing enhance the ability to track viral evolution, while international cooperation through organizations such as the World Organization for Animal Health (WOAH) and the Food and Agriculture Organization (FAO) facilitates data sharing and coordinated responses. Without comprehensive vaccine-oriented surveillance, immunization efforts may become ineffective due to the emergence of vaccine-resistant variants or undetected transmission routes. This AI strategy alongside stringent biosecurity measures and global cooperation, including geo-politically sensitive routes, is essential to mitigating the threat of HPAI and preventing future pandemics.