A study published in Radiology found that an off-label AI tool was effective in autonomously excluding pathology in chest X-rays with equal or lower rates of critical misses than radiologists, but AI mistakes were generally more clinically severe, suggesting the need for careful implementation.