The CDC placed early bets on AI — and now they are paying off

The CDC has quietly been building a modern AI infrastructure designed to reshape how public health data is collected, analyzed and acted upon.

• CDC
• Artificial Intelligence

When the White House unveiled the Genesis Mission, the sweeping governmentwide initiative to accelerate artificial intelligence using the combined power of national labs and high-performance computing, it was easy to see why the announcement grabbed headlines. Genesis promises to connect federal research data, supercomputers and scientific expertise into one massive discovery engine. The mission’s goal is nothing less than transforming how American science operates.

Genesis may be new, but the idea that federal agencies should use AI to solve entrenched problems is not. Some agencies have been laying groundwork for years, and one of the most advanced is the Centers for Disease Control and Prevention. While the public often associates the CDC with disease surveillance and outbreak response, the agency has quietly been building a modern AI infrastructure designed to reshape how public health data is collected, analyzed and acted upon.

The CDC published its vision for artificial intelligence back in 2023 as part of its Public Health Data Strategy. Even then, the agency recognized that AI could assist with everything from parsing laboratory data to automating public health reporting. Its focus was not on frontier AI models, but on practical systems that would help epidemiologists, state health departments and hospitals better understand what was happening on the ground.

As Nextgov/FCW reported at the time, the CDC’s broader modernization plan was aimed at improving interoperability, timeliness and data quality so responders would have actionable information when it mattered most.

That very early AI planning appears to be paying off. According to the recently updated Department of Health and Human Services AI Use Case Inventory, the CDC has already cited 54 instances where AI or machine learning has been deployed to support real work. Those projects span everything from anomaly detection to automated data extraction. A few of them stand out as clear examples of how much value AI is already adding to the public health mission.

One of the longest running programs to adopt AI is the National Syndromic Surveillance Program (NSSP), a nationwide collaboration that collects de-identified patient data from thousands of emergency departments. The program uses automated systems to scan incoming data for unusual patterns that may signal emerging outbreaks, environmental health events or injuries. NSSP’s near-real-time data feeds, combined with machine learning tools that help detect anomalies earlier and with greater accuracy, give state and local health departments a faster picture of what might be happening in their communities. 

Clever doctors no longer have to connect the dots, as the AI can alert them to common symptoms or unusual events clustered in specific areas. And the AI constantly does that now in near real time to enable medical teams to get ahead of outbreaks or other health emergencies.

Another prominent AI success recently shared by the CDC is called FluSight, a forecasting initiative that uses a variety of analytical approaches to predict influenza activity across the United States. Participating medical teams use both AI and machine learning models to combine traditional flu surveillance data with novel sources, like weather indicators and even social media trends. The AI is tasked with looking at data points that humans often overlook.

For example, if a cold and rainy spell in an area has traditionally led to more flu infections, then the risk of that can be raised if weather patterns in the future follow a similar trend. The CDC reports that these approaches have led to more accurate flu forecasts, which help providers and public health officials plan for surges, allocate resources and communicate early warnings when flu activity is expected to rise.

The CDC is also exploring AI tools that can help automate the labor-intensive process of scanning news reports for early signals of outbreaks. The agency notes that AI could be used to sort, categorize and summarize thousands of news articles a day as an additional window into pre-outbreak investigations. According to the CDC, that program is already processing roughly 8,000 articles daily, speeding up the event-based surveillance work that used to rely on staff manually identifying, reading and tagging relevant stories.

One of the clearest signs of the agency doubling down on its success came this fall with the launch of its AI Accelerator, a program designed to rapidly test new ideas and expand the ones that work. The accelerator is part of an effort to reduce the manual burden on epidemiologists and data personnel, who often deal with floods of unstructured information during outbreaks.

One of the most concrete examples of using AI as part of the Accelerator program involves computer vision, an advanced form of AI that can look at maps, photos and other visual media and make predictions based on what it sees. That is the basis for the CDC’s TowerScout program, which is designed to combat Legionnaires’ disease, which can hide in the aging water or cooling towers found in most cities.

According to the CDC, the traditional method of manually scanning images for disease-susceptible towers is slow and error-prone, especially across dense urban areas. TowerScout, by contrast, can spot cooling towers from aerial and satellite imagery in a fraction of the time. In fact, the AI tool reduces cooling-tower identification time from roughly four hours per area to just five minutes. This enables far faster source identification when outbreaks hit, and helps build registry databases for preventive maintenance and outreach.

The underlying AI model, developed originally at the UC Berkeley School of Information and now adopted by the CDC, identifies towers with high sensitivity and predictive accuracy, and has already supported dozens of outbreak investigations in multiple states.

What becomes clear in looking at all of these examples is that the CDC has been steadily applying AI where it can make the biggest difference in day-to-day public health work. None of the tools are flashy, but they solve real problems, save staff time and give health officials faster and better information when it matters most. As federal agencies look to build on emerging national AI initiatives, the CDC’s early and practical efforts offer a roadmap for how AI can strengthen critical missions long before the spotlight arrives.

John Breeden II is an award-winning journalist and reviewer with over 20 years of experience covering technology. He is the CEO of the Tech Writers Bureau, a group that creates technological thought leadership content for organizations of all sizes. Twitter: @LabGuys