Emerging Privacy Risks in AI: The Memorization Challenge in Language Models