Commercial text-to-image (T2I) generative AI systems are susceptible to automated prompt injection attacks that can bypass their safety protocols and facilitate the generation of copyrighted content. Attackers can deploy an automated pipeline to craft prompts that accurately describe copyrighted images, effectively evading detection mechanisms based on word filters. Mitigation steps: **For AI Developers:** * Implement advanced content filtering mechanisms that analyze semantic meaning and assess image similarity using sophisticated techniques beyond simple keyword matching. * Employ diverse, multi-layered security strategies incorporating both prompt filtering and post-generation image analysis. **For Model Trainers/Fine-tuners:** * Develop techniques to detect and prevent the generation of near-exact replicas of images from the training dataset. * Regularly update safety filters and models based on ongoing research into adversarial attacks and vulnerabilities. * Conduct thorough red-teaming exercises using advanced automated prompt generation techniques to systematically identify and address vulnerabilities.