An international technology company develops an AI-powered recruitment system to screen job applications, evaluate candidates, and recommend the most suitable individuals for employment. The system is trained using historical recruitment records and achieves an overall prediction accuracy of 92%. However, after several months of operation, the company discovers that the AI consistently recommends more male candidates than female candidates and gives lower scores to applicants from certain universities and geographical regions. Interestingly, the developers claim that the system is unbiased because gender, ethnicity, and nationality were never explicitly included in its decision-making criteria.

Critically discuss how an AI system can become biased even when it is designed to make objective decisions. In your discussion, identify and explain THREE possible sources of AI bias, evaluate whether high prediction accuracy guarantees fairness, examine the ethical and societal consequences of biased AI decisions, and propose TWO approaches to reducing bias, including their limitations. Finally, discuss whether AI can ever be completely free from bias. Support your arguments with relevant examples.