At a Glance
AI in public health includes machine learning for disease surveillance, natural language processing for outbreak detection, and generative AI tools for administrative and research tasks. The CDC reports 103 AI use cases as of December 31, 2025. Public health professionals with AI, data science, and informatics skills may be better positioned for emerging roles in agencies and health organizations.
The CDC says it had generated 103 AI use cases as of December 31, 2025. State and territorial health departments are catching up: ASTHO’s 2025 Profile data indicates AI policy adoption is growing, with statewide or agency-specific policies reported by many agencies. The field isn’t speculating about what AI might do for public health. It’s working through what AI is already doing and figuring out how to do it responsibly.
That shift happened fast. COVID-19 was one of the first pandemics managed with widespread digital surveillance, modeling, apps, and large-scale data systems, and it exposed both the potential and the limits of technology in a health crisis. Applications handled symptom reporting and contact tracing. Computer models simulated airflow in outbreak centers. Data systems that had taken years to build were pushed to their limits overnight. Public health professionals who came through that period produced a clear list of what technology needed to do better the next time around.
Generative AI, which began to advance rapidly as the pandemic was still underway, has become one of the most consequential additions to that toolkit. Understanding how it works, where it helps, and where it creates risk is no longer optional for anyone working in the field.
Machine Learning and AI in Public Health Surveillance
Machine learning has been part of public health data work for more than a decade. When it comes to analyzing large datasets from disease surveillance (case reports, lab results, death records, prescription data), ML has been a reliable tool for identifying patterns that human analysts might miss or take longer to find.
The CDC has put this to work in specific, documented ways. An ML system parses death certificate language to capture opioid-related deaths more accurately, catching variant terms and misspellings that cause undercounts in manual review. AI tools can help screen chest X-rays for tuberculosis and may reduce radiologist workload, though results generally require clinical validation and oversight. TowerScout, a tool developed with Los Angeles County, uses computer vision to identify cooling towers in satellite imagery, a key step in Legionnaires’ disease outbreak preparedness that would be impractical to do by hand across a county of that size.
Several of these use cases are documented by CDC and HHS and are being piloted, deployed, or scaled in public health settings. The CDC’s AI strategy describes accelerated AI adoption across public health workflows as a named priority, with the agency’s AI Accelerator Program (AIX) serving as a model for piloting and scaling high-impact solutions.
Generative AI Enters the Picture
Machine learning and generative AI are different tools, though both fall under the broader umbrella of artificial intelligence. Traditional ML excels at pattern recognition in structured datasets. Generative AI, including large language models (LLMs), is built to process and produce language, summarize documents, answer questions, and draft text.
Some state agencies report using generative AI tools for day-to-day work. The CDC’s 2025 Epidemiology and Laboratory Capacity survey found that state agencies report using these tools to draft and edit reports, summarize meeting recordings, generate or correct code for data tasks, and run pilot chatbots to handle common public inquiries. Several agencies have also used AI-assisted redaction to speed up privacy review of documents, keeping humans in the loop while cutting the time spent on routine screening.
One CDC tool, NewsScape, uses an LLM to scan approximately 8,000 news articles and public health reports per day, flagging potential health threats for case-based and event-based surveillance teams. At that volume, the work of filtering signal from noise would be out of reach for a human team alone.
What AI Can Do in Public Health
The applications of AI in public health span the full arc of the field, from early detection to prevention to communications.
Disease Detection and Outbreak Response
AI systems can connect data sources that traditional surveillance methods treat separately. Social media posts, electronic health records, pharmacy data, and environmental sensors can be integrated and scanned in real time. ASTHO reports that a small number of states, including Minnesota and California, use AI-enabled real-time disease surveillance systems that can identify outbreaks weeks before traditional systems. Earlier warning can support faster investigation and response, though containment also depends on staffing, data quality, public cooperation, and available interventions.
During an active outbreak, the speed advantage compounds. AI can process incoming case data, map transmission networks, and model likely spread while public health teams are still receiving reports manually. Algorithms can surface connections between patients, a shared location or a common exposure window, that human investigators might reach eventually but not quickly enough.
NLP and Public Health Communications
Natural language processing (NLP) gives public health agencies a way to handle communication volume that would otherwise overwhelm their staff. Phone hotlines and emergency rooms fill up fast when a new health threat surfaces. Misinformation spreads in that vacuum.
NLP-powered chatbots and voice systems can handle basic symptom screening, answer frequently asked questions, and gather contact information for follow-up. They work in multiple languages and don’t require a public health professional on the other end. The data they collect also feeds into real-time analytics, helping officials see where questions and concerns cluster geographically before they become a crisis.
Prevention and Chronic Disease
Prevention is where AI’s reach extends furthest beyond infectious disease. The CDC has been developing AI models to forecast trends in opioid overdose mortality by tying together datasets from across the country. That kind of predictive work can direct resources to communities before mortality spikes, rather than in response to them.
In the clinical setting, AI tools that scan electronic health records for early risk indicators can flag patients who need preventive intervention before a condition progresses. Population-level screening at that scale wasn’t feasible before machine learning made it computationally practical.
Drug Discovery and Emergency Preparedness
AI is increasingly used in drug discovery and repurposing research, and its real-world impact on emergency pharmaceutical timelines varies by disease, product, regulatory requirements, and trial evidence. Tools that screen chemical compounds for development potential and identify existing drugs with possible new applications are active areas of research at major health institutions.
Risks and Considerations
Using AI well in public health requires understanding where it fails, not just where it succeeds. There are several categories of risk that any public health professional working with AI systems needs to be aware of.
Bias in Training Data
AI systems learn from historical data. Healthcare data collected over decades reflects the discrimination and inequities of those eras. Populations that were underserved by the healthcare system are also underrepresented or misrepresented in the datasets that train these models. That bias doesn’t disappear when the data is fed into an algorithm. It can get amplified.
In public health terms, a biased model isn’t just an equity problem. It’s an accuracy problem. A disease surveillance system that performs poorly on data from certain communities will miss outbreaks in those communities, and an unchecked outbreak anywhere is a threat everywhere. Algorithmic fairness in public health tools is a technical requirement, not just an ethical aspiration.
Hallucinations and LLM Reliability
Large language models generate text based on statistical patterns in their training data. They don’t have access to verified facts and can produce confident-sounding, incorrect statements. In a clinical or public health communication context, that’s a meaningful risk. Any LLM-generated content intended for public consumption or clinical decision support needs human review before it goes out.
Privacy
Public health surveillance has always required digging into individual behavior and health history. AI amplifies that reach considerably. Weaving together social media, health records, location data, and purchasing patterns can reveal details people haven’t chosen to disclose. The AIDS epidemic showed how epidemiological work can inadvertently out people when privacy protections aren’t built into the process from the start. The same risk exists in AI-driven surveillance at higher speeds and on a larger scale.
HIPAA governs protected health information, and LLMs trained on or processing patient data represent a potential compliance risk. Building privacy protections into AI systems from the design stage is a public health governance obligation, not an afterthought.
Explainability
Some of the most powerful ML models are black boxes. Their outputs can be accurate and useful, but the internal reasoning that produced them isn’t visible to the researchers or officials relying on them. For public health decisions that affect policy, resource allocation, or individual clinical care, that lack of explainability is a problem. Results need to be checked. Recommendations from AI systems should be treated as inputs to human judgment, not substitutes for it.
Governance and Policy Frameworks
The policy landscape around AI in public health is developing alongside the technology itself. HHS released an AI Strategy on December 4, 2025, focused on expanding AI use across internal operations, research, and public health efforts. The CDC’s Public Health Data Strategy, updated annually, includes AI integration as a named priority, with goals focused on responsible deployment at scale and the reduction of manual burden in state and local health agencies.
At the state level, ASTHO’s 2025 Profile data shows uneven adoption of AI governance, with statewide and agency-specific policy approaches reported across many agencies. But a significant share of agencies are still using consumer-grade AI tools for non-sensitive work without enterprise-level governance. The gap between agencies with mature AI infrastructure and those still in the early stages of adoption is wide.
That gap matters for public health outcomes. AI tools that improve disease surveillance or reduce administrative burden in states with strong adoption aren’t available to populations in states without it. Equity in AI adoption is becoming an equity issue in public health capacity.
Getting Educated for AI in Public Health
Degree programs are adapting, though at different speeds. Accredited MPH programs commonly include exposure to data science, biostatistics, and health informatics, the foundational disciplines that have always intersected with machine learning in surveillance and research. Those foundations are more relevant now than they’ve ever been.
For professionals who want deeper technical preparation, specialized programs are available. A Master of Science in Health Informatics or a graduate certificate in healthcare analytics can build the skills to work directly with AI systems in a public health context. Some programs now offer specific coursework in computational epidemiology, which sits at the intersection of data science, ML, and disease modeling.
For working professionals who don’t need a full degree, short certificate programs in AI, health analytics, and healthcare data science are increasingly available. They cover enough on ML techniques, data analysis, and AI ethics to give practitioners a working understanding of the tools their agencies are already using.
What all of them require, beyond the technical content, is enough grounding in ethics, privacy, and governance to know when to push back on an AI recommendation and when to trust it. That judgment doesn’t come from the algorithm. It comes from the professional.
Frequently Asked Questions
How is AI currently being used in public health agencies?
Federal and state agencies are using AI for disease surveillance, administrative document review, outbreak modeling, and public communications. The CDC reports 103 AI use cases as of December 31, 2025, including tools that scan thousands of health news articles daily for threat signals, identify cooling towers in satellite images for Legionnaires’ preparedness, and parse death certificate data to improve opioid mortality counts. Some state agencies report using generative AI tools to draft reports, summarize meeting notes, and handle routine public inquiries via chatbots.
What is the difference between machine learning and generative AI in a public health context?
Machine learning in public health typically refers to algorithms trained on structured datasets to identify patterns, make predictions, or classify information. It’s been used for years in disease surveillance and health records analysis. Generative AI, including large language models, is a newer category that processes and produces language. In public health, it’s being applied to document summarization, communication, and literature synthesis, but it requires more oversight because it can confidently produce incorrect information.
What are the biggest risks of using AI in public health?
Bias in training data is the most pervasive risk. Healthcare data collected over decades reflects historical inequities, and models trained on that data can underperform for underrepresented populations. LLM hallucinations pose a risk in any context where accuracy is critical. Privacy exposure is a concern when AI systems integrate multiple data sources about individuals. And the opacity of some advanced models creates challenges for the scientific accountability that public health decisions require.
Do public health professionals need to learn to code to work with AI?
Not necessarily. Many AI tools in public health are used through interfaces that don’t require programming. But a working understanding of how the tools function, what data they rely on, and where they can go wrong is increasingly expected of professionals at all levels. Epidemiologists, health officials, and program administrators who understand the basics of data science are better positioned to evaluate AI outputs critically and advocate for responsible use.
What degree is best for a career at the intersection of AI and public health?
An MPH with a concentration in epidemiology, biostatistics, or health informatics provides a strong foundation. For roles with significant technical responsibility, a master’s degree in health informatics, data science, or computational epidemiology offers more depth. Certificate programs in AI for healthcare are a practical option for working professionals who need to update their skills without completing another degree.
Key Takeaways
- AI in public health is already operational, not theoretical. The CDC reports 103 AI use cases as of December 31, 2025, and ASTHO’s 2025 data shows AI policy adoption growing across state and territorial health agencies.
- Machine learning and generative AI are different tools with different risk profiles. ML excels at pattern recognition in structured data. LLMs handle language but require careful human oversight to avoid misinformation.
- Bias in training data is both an equity problem and an accuracy problem. A surveillance system that misses trends in underrepresented populations is a public health risk, not just an ethical failure.
- Federal policy frameworks from HHS and CDC are developing, but adoption is uneven at the state level. The gap in AI capacity between well-resourced and less-resourced agencies is becoming a public health equity issue.
- Public health education is adapting. MPH programs, specialized master’s degrees, and certificate programs all offer pathways to build the technical and ethical grounding AI-era public health work requires.
Looking to build the skills for a career in public health? Browse accredited degree programs, from MPH to health informatics, and find options that fit your goals.
Find Schools Near You
Laura Bennett, MPH is a public health professional with over 12 years of experience in community health education and program coordination. She specializes in helping aspiring professionals explore flexible education pathways, including online and hybrid public health degree programs. Laura is passionate about making public health careers more accessible through practical, accredited training
