HHS Developing AI Tool to Analyze Vaccine Safety Data: Concerns Rise Under Kennedy’s Leadership

The Department of Health and Human Services is developing a generative AI tool to analyze data from the Vaccine Adverse Event Reporting System (VAERS), raising concerns about how it might be used under the leadership of Secretary Robert F. Kennedy Jr., a known vaccine skeptic.

The Controversial AI Initiative

According to a recently released inventory of AI applications planned for 2025, HHS has been developing a generative artificial intelligence tool since late 2023 that would search for patterns in vaccine monitoring data and generate hypotheses about potential negative effects of vaccines.

This development comes amid significant changes to vaccine policies under Kennedy’s leadership, including the removal of several vaccines from the recommended childhood immunization schedule, such as those for COVID-19, influenza, hepatitis A and B, meningococcal disease, rotavirus, and RSV.

Understanding VAERS and Its Limitations

VAERS, established in 1990 and jointly managed by the CDC and FDA, was designed as an early warning system to detect potential safety issues with vaccines after approval. However, the system has significant limitations:

• Anyone can submit reports to VAERS without verification
• Reports show correlation (events following vaccination) but not causation
• The database lacks control group data
• It doesn’t include information on how many people received a vaccine, making risk assessment difficult

Dr. Paul Offit, director of the Vaccine Education Center at Children’s Hospital of Philadelphia, describes VAERS as “a hypothesis-generating mechanism” and “a noisy system” that doesn’t prove vaccines caused reported reactions.

AI Applications and Concerns

While government scientists have used traditional natural language processing to analyze VAERS data for years, the shift to more advanced large language models (LLMs) raises new concerns:

• LLMs are known for producing convincing but false information (hallucinations)
• Human experts must validate any AI-generated hypotheses
• With CDC staffing cuts, questions remain about capacity to properly investigate AI findings

Leslie Lenert, formerly of the CDC and now at Rutgers University, emphasizes the importance of pairing VAERS information with other data sources to determine actual risks.

Political Context and Potential Misuse

Experts worry that the AI tool could be misused to support Kennedy’s anti-vaccine agenda. Kennedy has proposed overhauling VAERS, claiming it suppresses information about vaccine side effects, and has suggested changes to the Vaccine Injury Compensation Program that could make it easier to sue for unproven adverse events.

These concerns are heightened by recent developments at the FDA, where Vinay Prasad, director of the Center for Biologics Evaluation and Research, reportedly proposed stricter vaccine regulation based on VAERS reports that had already been reviewed, prompting pushback from former FDA commissioners.

Balancing Innovation and Safety

Despite concerns, experts acknowledge potential benefits of using advanced AI to analyze vaccine safety data. Jesse Goodman of Georgetown University notes that LLMs could potentially detect previously unknown safety issues, but emphasizes the need for thorough human investigation of any findings.

In the past, VAERS has successfully flagged legitimate safety concerns, including rare clotting disorders associated with the Johnson & Johnson COVID-19 vaccine and cases of myocarditis following mRNA COVID-19 vaccines, demonstrating its value when properly used and interpreted.