MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed “PhotoGuard,” a technique that uses perturbations to protect images from manipulation by AI models. This new measure is necessary as advanced generative models like DALL-E and Midjourney have made it easier for even inexperienced users to manipulate images. Misuse of these technologies can have serious consequences, from market manipulation to personal image blackmail. PhotoGuard disrupts the AI model’s ability to manipulate images by introducing minuscule alterations that are invisible to the human eye but detectable by computer models. There are two attack methods used in PhotoGuard: the “encoder” attack, which targets the image’s latent representation, and the “diffusion” attack, which optimizes perturbations to make the final image resemble a target image. This technique preserves the image’s visual integrity while ensuring its protection. While PhotoGuard is effective, it’s not a complete solution, and a collaborative approach involving model developers, social media platforms, and policymakers is needed to combat unauthorized image manipulation. Developers should invest in engineering robust immunizations against potential threats posed by AI tools. The paper was presented at the International Conference on Artificial Intelligence and Statistics (AISTATS).