A researcher at Arizona State University is leading efforts to create stronger standards for identifying and managing AI-generated images and other synthetic media. The article highlights the work of Yezhou “YZ” Yang, an associate professor in computer science and engineering, who is helping bring together researchers, technology companies, and policy groups to build a global framework for AI transparency, privacy, and ethical safeguards. The need for such standards is becoming urgent as AI-generated visuals are now almost indistinguishable from real content.
A major focus of the work is the development of digital watermarking and detectable signals embedded directly into AI-generated content. These hidden markers act like digital fingerprints that can later be identified by verification systems. The article notes that studies have found people can correctly distinguish AI-generated media from real images only about 51 percent of the time, which is almost the same as random guessing. This highlights the growing challenge of misinformation, deepfakes, and fraudulent media in the digital space.
Another important aspect of the research is machine unlearning, which allows AI systems to selectively “forget” specific data, concepts, or harmful content without retraining the entire model. This can be especially useful for removing copyrighted material, sensitive personal information, or biased outputs after a system has already been deployed. Yang’s team is working on advanced methods that can erase unwanted concepts while still maintaining the overall quality and performance of the AI model.
Overall, the article emphasizes that the challenge is no longer only about detecting fake content, but about preserving trust in digital information ecosystems. By combining detection standards, watermarking, and machine unlearning, this research aims to create a future where AI-generated media is both transparent and accountable, helping society respond more effectively to the risks of deepfakes and misinformation.
