Unexpected capabilities or behaviors that arise from AI systems beyond their explicit programming. Occurs when complex systems exhibit properties that cannot be predicted from understanding individual components. Can be beneficial or concerning, representing both AI advancement and potential risks.