As artificial intelligence technology continues to advance rapidly, the ability to generate and discern deepfake content will become a significant societal challenge. Recent studies indicate that the digital watermarking technology currently in wide use can be easily circumvented, and the difficulty in preventing AI-generated deepfakes remains high. Research teams have successfully demonstrated the vulnerability of existing watermark protection measures, warning that the misuse of AI technology could cause widespread harm and must be taken seriously. Although difficulties persist, designing a reliable digital watermarking system is not entirely out of reach. Experts from various fields are calling for stronger collaboration between industries and government departments in preventing AI-generated deepfakes, and for the cautious application of emerging technologies to mitigate potential risks.