With the advancement of generative AI, such as DALL·E and others, creating hyper-realistic images has become increasingly accessible. This brings creative benefits but also significant risks, such as image falsification used to deceive people and manipulate information.
Currently, regulations and guidelines require AI-generated content to be labeled, but these solutions rely on the good faith of developers and human laws, which can be circumvented. The question I want to raise is:
What if AI itself were designed to ensure its images are always identifiable in an inviolable way?
I propose a native, encrypted, and dynamic system integrated into AI image generators, operating under the following principles:
Encrypted and Dynamic Watermarking – A mandatory digital signature embedded in the image that automatically changes to prevent removal.
Unalterable Code of Conduct – The AI’s core should prevent humans from removing or tampering with this identification, ensuring no image can be used to deceive people as if it were a real photograph.
Multi-Layer Authentication – Utilizing asymmetric cryptography and potentially blockchain to ensure any trusted entity can verify the image’s origin.
Protection Against Reverse Engineering – The system should renew itself periodically to prevent code manipulation and ensure continuous traceability.
Discussion Points
- Could this approach be implemented without compromising creativity and privacy?
- Would using blockchain to track image origins be viable on a large scale?
- How can we ensure this solution is not used to restrict the legitimate use of AI for art and expression?
My intention in presenting this idea is to open a discussion on how we can ensure AI-generated images remain traceable, preventing dangerous deepfakes without stifling innovation so that we can use technology to strengthen and build truth, rather than amplifying our flaws and allowing it to be wielded in a dishonest and destructive way.
I’d love to hear thoughts and suggestions on this proposal!