Deepfakes in the AI era: why watermarks are not enough

Our comments on the article

• “Watermarks alone will not be enough. They are easy to remove, as is metadata, especially for those interested in spreading disinformation.”

• “Developers should be required to mark generated content with cryptographic tags, use technologies such as blockchain and hash identifiers, and platforms should verify and label such material—or respond promptly to complaints.”

• “Platforms are already required to restrict access to prohibited information. This principle can be extended and codified for AI-generated content.”