A software developer claims to have successfully reverse-engineered Google DeepMind’s SynthID system, demonstrating how AI watermarks can be removed from generated images or inserted into other content. Google, however, refutes this claim, stating that the system remains secure.
Source: The Verge AI