When an AI system contributes to a medical error, hospitals face steeper public backlash than when a human physician alone makes the same mistake. A new study published in Nature found that people attribute significantly more responsibility to hospitals following adverse events involving artificial intelligence, and they're more likely to file complaints and pursue legal action. However, the way physicians and AI systems work together matters enormously: when doctors actively collaborate with AI rather than relying on it independently, public trust in the hospital remains substantially higher.

Why Does AI Involvement Change How People Judge Medical Errors?

Researchers at Nature conducted two separate studies to understand public reactions to AI-involved adverse events in healthcare. In the first study, nearly 300 participants reviewed hypothetical scenarios involving a missed cancer diagnosis during a capsule endoscopy, a procedure that uses a swallowable camera to examine the digestive tract. The researchers varied who made the initial "normal" diagnosis: an AI system alone, a human endoscopist alone, or both working together with the AI's interpretation reviewed by the physician.

The results were striking. When participants learned that the diagnosis had missed small bowel adenocarcinoma five months later, their responses differed dramatically based on who was responsible. Participants attributed significantly more responsibility to the hospital when AI was involved, either alone or jointly with a physician, compared to when a human physician made the error independently. The difference was even more pronounced when AI made the diagnosis without any physician review afterward.

This pattern held across all three measures the researchers tracked: perceived hospital responsibility, likelihood of filing a complaint, and likelihood of pursuing legal action. The findings suggest that the public views hospitals differently when AI is part of the decision-making chain, perhaps because organizations are seen as responsible for choosing to deploy and oversee the technology.

How Does Physician-AI Collaboration Change the Picture?

The second study dug deeper into a crucial finding from the first: when physicians actively reviewed AI outputs, public reactions became less negative. Researchers tested different types of collaboration between doctors and AI systems to see which approaches best preserved public trust.

The key insight is that not all physician involvement is equal. Interactive collaboration, where physicians and AI systems work together throughout the diagnostic process rather than sequentially, significantly reduced negative public reactions compared to AI working alone. This suggests that transparency about how AI and humans interact in medical decision-making may help hospitals maintain trust even when errors occur.

"Although AI is increasingly implemented in healthcare, little is known about how the public assigns accountability to hospitals following AI-involved adverse events," the researchers noted in their study.

Nature Study Authors

The implications are substantial. As hospitals adopt AI tools at an accelerating pace, they face a paradox: AI systems promise to reduce operational burdens and improve clinical performance, yet adverse events involving AI may carry reputational, financial, and legal costs that could undermine those gains. The study found that approximately 17,000 malpractice lawsuits are filed annually in the United States, totaling roughly $4 billion in payouts to patients, and AI-involved errors could shift how juries and the public assign blame.

How to Build Public Trust in AI-Assisted Medical Care

• Emphasize Physician Oversight: Hospitals should clearly communicate that physicians actively review and collaborate with AI systems rather than relying on AI recommendations passively. Interactive collaboration throughout the diagnostic process, not just at the end, appears to preserve public trust more effectively.
• Implement Transparent Workflows: Design clinical workflows where the role of AI is visible and understood by patients and families. When people understand how AI supports rather than replaces physician judgment, their confidence in the hospital increases.
• Establish Clear Accountability Structures: Develop organizational protocols that make it clear how hospitals oversee AI deployment, monitor performance, and intervene when needed. Public perception of hospital responsibility for AI systems is high, so governance must be robust and visible.

What Does This Mean for Radiology and Other AI-Heavy Specialties?

Radiology is one of the first medical specialties to embrace AI at scale, with systems now helping interpret everything from chest X-rays to mammograms. A new AI-native reporting platform called Mosaic Reporting, launched by Mosaic Clinical Technologies in June 2026, illustrates how the industry is evolving. The system uses artificial intelligence to assist radiologists in real time as they interpret medical images, automatically organizing findings and structuring reports while keeping the radiologist in control of clinical decision-making.

Mosaic Reporting is already deployed to thousands of radiologists through Radiology Partners-affiliated practices, providing a real-world testing ground for how AI can support rather than replace physician expertise. The platform is powered by foundation models developed by Cognita Imaging, the AI division of Mosaic Clinical Technologies, and offers improved speed and accuracy compared to third-party models.

"Radiology is facing a chronic shortage of radiologists as imaging volumes continue to climb, leaving radiology practices, departments and the health systems that depend on them with growing backlogs and turnaround time pressure," said Mike Peresie, president of Mosaic Clinical Technologies. "Mosaic Reporting is designed to help healthcare organizations create reading capacity, improve turnaround times and better meet service levels that referring physicians and patients expect."

Mike Peresie, President of Mosaic Clinical Technologies

The platform's design reflects lessons from the Nature study: it keeps radiologists at the center of clinical decision-making while AI handles administrative and organizational tasks. Rather than making diagnoses independently, the system assists with real-time report construction, intelligent extraction of findings from natural speech, and targeted editing capabilities. This collaborative approach, where AI reduces cognitive burden rather than replacing judgment, aligns with what the research shows builds public trust.

"Most AI documentation tools are built around clinician-patient conversations, but radiologists interpret images, which demands a fundamentally different approach," explained Adrit Rao, technical product lead at Cognita Imaging. "Mosaic Reporting is one of the first new and significantly advanced approaches to radiology reporting in more than a decade. We worked closely with radiologists to design a reporting experience where AI works in real time as they read, with targeted edits and structured output that reduce friction and keep clinical decision-making at the center."

Adrit Rao, Technical Product Lead at Cognita Imaging

For healthcare organizations deploying AI, the message is clear: the technology's success depends not just on technical performance but on how it's integrated into clinical workflows and communicated to patients and the public. As AI becomes more embedded in medical practice, hospitals that prioritize transparent, collaborative approaches where physicians remain visibly in control are likely to maintain patient trust even when errors occur. The alternative, where AI operates as a black box or makes recommendations without physician review, carries significant reputational and legal risks that could offset the operational benefits the technology promises to deliver.