OpenAI acknowledges that these watermarks are not foolproof and can be easily circumvented through paraphrasing, image editing, or simple re-prompting techniques.
Open the full topic