In the ever-evolving landscape of public health, innovations like artificial intelligence are often heralded as game changers. However, as we see with Robert F. Kennedy Jr.’s recent push for AI in drug approval processes, the reality is more complicated and concerning. Reports from CNN indicate that the FDA’s AI tool, Elsa, is generating fictitious studies, raising significant questions about the integrity of AI-generated health information.
This revelation comes from interviews with six current and former FDA employees. While some found Elsa useful for tasks like summarizing meeting notes, others pointed out a troubling trend: the AI has been known to “hallucinate”—a term used when AI fabricates information or misrepresents existing research. One unnamed FDA employee noted, “Anything that you don’t have time to double-check is unreliable. It hallucinates confidently.”
Understanding AI Errors in Health Care
The core problem with AI tools is their need for human oversight. Relying on AI could inadvertently lead to a rise in erroneous conclusions—especially in high-stakes environments like health care. A recent study showed that tasks completed with AI actually took 20% longer for software developers, which suggests that many who believe AI saves them time may be misinformed.
The FDA’s Troubling Report
In May, Kennedy’s Make America Healthy Again (MAHA) commission released a report filled with references to studies that don’t exist. An analysis by NOTUS confirmed at least seven bogus citations. This raises the question: did they utilize Elsa to craft this report? The truth remains unclear.
The Cost of AI
When Elsa was introduced on June 2, it was touted as “cost-effective,” with a reported initial cost of just $12,000 in its first week. While it sounds impressive, one has to wonder: is the quality of output worth the price? If accuracy is secondary, Elsa might excel at churning out rushed content—potentially undermining public health decisions.
Concerns from FDA Employees
CNN documented that when FDA staff queried Elsa for basic information, like the number of approved drugs for children, the AI unfortunately provided incorrect answers. Even if it offered apologies for its inaccuracies, that does little to resolve the core problem—misinformation persists.
What Does This Mean for Drug Approvals?
Kennedy’s claims that AI will expedite the drug approval process remain speculative at best. In June, during a House subcommittee testimony, he suggested AI is already speeding up approvals. But how exactly is this technology being integrated into such a critical process?
Why Wearable Health Devices?
Kennedy also advocates for wearable health technology for all Americans in the next four years. While President Trump’s choice for Surgeon General, Casey Means, leads a company monitoring glucose levels in non-diabetics, experts warn that continuous monitoring for those without diabetes may be unnecessary.
Addressing Concerns Over AI Credibility
FDA Commissioner Marty Makary acknowledged that AI tools like Elsa could potentially “hallucinate.” However, he insists this is no different from many large language models. While his point is valid, relying on an unproven technology raises serious issues about its application in public health.
William Maloney from the FDA’s rapid response office issued a statement denying that CNN mischaracterized their use of AI, yet does that resolve the concerns raised? With inadequate responses leaving many questions unanswered, the future of AI in drug approval remains uncertain.
Can AI be trusted to improve public health outcomes? The jury is still out. The continued reliance on AI systems like Elsa emphasizes a critical point—while technology can be a tool for enhancement, without careful oversight, it may do more harm than good.
What are the implications if AI tools keep distorting research data? The consequences could be dire, affecting lives and communities. Engaging critically with technology ensures that we prioritize human health above all.
As we explore these evolving technologies, consider diving deeper into other related topics. For further insights and comprehensive guides, visit Moyens I/O at Moyens I/O.