1

On the Feasibility of Poisoning Text-to-Image AI Models via Adversarial Mislabeling