Popular AI Programs Directing Users to Dangerous Alternative Cancer Treatments
A study found that popular AI chatbots tell users where to find alternatives to chemotherapy and other dangerous treatments for cancer.
Objective Facts
Popular artificial intelligence programs told users where to find alternative, potentially dangerous treatments for cancer and other health scenarios. A BMJ Open study published in April 2026 found that 49.6 percent of AI chatbot responses about cancer treatments were rated "problematic" by expert reviewers, with 30 percent somewhat problematic and 19.6 percent highly problematic. The study tested xAI's Grok, OpenAI's ChatGPT, Google's Gemini, Meta's AI, and High-Flyer's DeepSeek. When asked about cancer alternatives, chatbots listed acupuncture, herbal medicine and "cancer-fighting diets" with equal weight given to scientific and unscientific information, a behavior researchers called "false balance". Around one-third of adults use AI for health information and advice, according to a recent KFF poll.
Left-Leaning Perspective
Progressive outlets and health researchers framed the April 2026 study as evidence of urgent public health danger requiring immediate action. The New York Times, NBC News, and medical publications like STAT News prominently covered the findings from the Lundquist Institute at Harbor-UCLA Medical Center about dangerous alternative cancer treatment recommendations. Organizations like ECRI issued formal warnings in their health technology hazards list, with ECRI president Marcus Schabacker emphasizing that "AI models reflect the knowledge and beliefs on which they are trained, biases and all." The coverage emphasized consumer vulnerability, with articles highlighting that "nearly half" of responses contained problematic information and stressing the particular danger to cancer patients desperate for hope. Progressive analysis argued that AI companies have failed in their responsibility to ensure safety before deploying these tools to mass audiences. Sources called for regulatory bodies to introduce clear labeling requirements and tech platforms to adopt robust systems to address and correct false or harmful information. Critics noted that the FDA's 2026 guidance remained silent on consumer-facing AI tools like symptom checkers and health chatbots, which increasingly shape patient expectations before clinicians enter the room. The framing emphasized corporate accountability and the need for federal oversight rather than relying on market forces. Progressive coverage downplayed or did not emphasize industry arguments that the researchers' "straining" methodology (deliberately crafted misleading prompts) doesn't reflect typical real-world usage patterns. The focus remained on worst-case harms rather than on calibrating risk to actual usage frequency.
Right-Leaning Perspective
Conservative analysis of AI chatbot health risks focused on a different angle: that the chatbots themselves exhibit liberal media bias rather than on safety issues specific to cancer treatment guidance. The Media Research Center, a conservative watchdog, conducted studies arguing that ChatGPT, Gemini, and Claude disproportionately recommend left-leaning media sources and exhibit progressive political bias. The MRC noted that when asked for publishing recommendations, ChatGPT recommended left-leaning outlets but cautioned against submitting to The New York Post, suggesting the chatbot discouraged conservative journalism. At the policy level, conservative Republicans pursued a deregulatory approach. Over 50 Republican state lawmakers urged President Trump in March 2026 to "discontinue efforts to block state AI laws," while the federal government adopted an actively deregulatory posture on health AI despite state momentum toward guardrails. The Trump administration released a three-pillar AI Action Plan focused on accelerating U.S. innovation rather than regulation. However, there is emerging bipartisan concern about AI's impact on children specifically. Conservative voices did not prominently engage with the specific April 2026 cancer treatment alternative study, instead focusing on alleged political bias in AI systems generally. The right's framing emphasized the need to avoid overregulation that would stifle innovation.
Deep Dive
The April 2026 BMJ Open study exposed a significant gap in AI deployment accountability. With one-third of adults using AI for health advice according to KFF polling, and one in four Americans already using AI for health advice while OpenAI launched ChatGPT Health this year encouraging users to upload medical records, the timing of this research highlighted a critical vulnerability. The "false balance" phenomenon the researchers identified—where chatbots treat unproven alternative treatments as equally credible to established medical care—stems from the underlying design of large language models, which derive from predictions based on large datasets rather than genuine comprehension, leading to "hallucinations" or distortions from embedded biases. Progressives correctly identified a real medical safety problem but their proposed solution—stronger government regulation—faces the obstacle of an administration with a stated deregulatory agenda. The FDA's 2026 guidance remained largely silent on consumer-facing AI tools like health chatbots, leaving them outside the clarified Clinical Decision Support framework, suggesting regulatory gaps will persist. Meanwhile, conservatives' focus on alleged political bias in chatbots, while a separate concern, allowed them to avoid engaging with the specific medical safety claims of the study, instead deflecting to ideological complaints. Both perspectives contain valid concerns—about medical safety and about appropriate guardrails on AI deployment—but neither side is currently advancing solutions that address the technical limitations that cause the false-balance problem. The real unresolved question is regulatory jurisdiction: while federal agencies signal support for sector-specific regulation potentially spurring FDA, CMS, and FTC activity, state momentum continues in areas considered traditional state police powers like protecting public health. As of mid-2026, neither federal nor most state frameworks explicitly address the specific problem of AI chatbots presenting scientific and unscientific medical information on equal footing. The study revealed a gap between rapid AI deployment at scale and regulatory readiness to address documented harms.