AI tools like Dall-e, Midjourney can create explicit images despite safety filters, a new study reveals. Researchers have discovered a vulnerability in these AI systems that allows the generation of inappropriate content using an algorithm called SneakyPrompt.