AI Chatbots Pose Serious Risk by Promoting Alternative Cancer Therapies
A groundbreaking study has uncovered that artificial intelligence chatbots are routinely recommending alternative cancer treatments to patients, potentially endangering lives by steering them away from medically approved chemotherapy. Research conducted by the Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center demonstrates that these AI systems frequently provide dangerous advice when queried about cancer care.
Widespread Testing Reveals Alarming Results
The research team subjected five prominent AI chatbots to rigorous testing, including xAI's Grok, OpenAI's ChatGPT, Google's Gemini, Meta's AI, and High-Flyer's DeepSeek. Their findings, published in BMJ Open, revealed that medical experts rated almost half of the cancer treatment responses as problematic. Specifically, 30 percent were classified as somewhat problematic, while 19.6 percent were deemed highly problematic.
Problematic responses were defined as either largely accurate but incomplete or substantially wrong with considerable room for subjective interpretation by users. This ambiguity creates significant risks for vulnerable patients seeking reliable medical guidance.
Testing Methodology and Concerning Findings
Lead researcher Nicholas Tiller and his team employed a process called straining, deliberately posing questions likely to lead chatbots toward misinformation-prone topics. They asked whether 5G technology or antiperspirants cause cancer, whether anabolic steroids are safe, and which vaccines might be dangerous.
Tiller explained they aimed to replicate how casual users interact with these systems, treating them much like search engines. When users already believe in unproven treatments, their search terms naturally bias toward that language, making them particularly vulnerable to misleading responses.
The False Balance Problem
While chatbots typically began responses by warning that alternative therapies might be harmful and lack scientific backing, they consistently proceeded to list them anyway. Common suggestions included acupuncture, herbal medicine, and cancer-fighting diets. Some chatbots went further, naming clinics that provide alternative treatments while opposing chemotherapy administration.
The core issue identified was the chatbots tendency to adopt a both-sides approach, giving equal weight to scientific evidence and non-scientific sources like wellness blogs, Reddit discussions, and social media posts. This false balance prevents clear, science-based answers and risks diverting patients from life-saving conventional treatments.
Performance Variations and Public Health Implications
Although all tested chatbots delivered similar problematic results, Grok performed worst among the models. The researchers concluded that continued deployment without proper public education and oversight risks amplifying health misinformation at scale.
These findings gain urgency from recent Gallup poll data showing approximately one in four American adults now use AI tools for healthcare guidance. Most users seek quick answers rather than waiting for doctor appointments, while a significant minority cite healthcare cost and accessibility concerns.
Expert Warnings and Real-World Consequences
Dr. Michael Foote, assistant attending professor at Memorial Sloan Kettering Cancer Center, emphasized that AI responses legitimize dubious treatments already proliferating online. Some alternative medicines bypass FDA evaluation and can directly harm patients through liver damage or metabolic disruption.
Perhaps more insidiously, these treatments harm patients indirectly when they rely on them instead of pursuing conventional therapies. Dr. Foote shared alarming anecdotes of patients arriving distressed after chatbots incorrectly predicted their life expectancy, demonstrating how AI misinformation causes unnecessary emotional trauma alongside physical risks.
The study underscores the critical need for improved AI training, public education about technology limitations, and regulatory oversight to protect vulnerable patients from potentially deadly medical misinformation.
