04:00
2026-05-28
arxiv.org
ai-safety
PAST2HARM: A Simple Adaptive Past Tense Attack for Jailbreaking Multimodal AI
A new jailbreak framework called PAST2HARM has achieved up to 100% attack success rates against state-of-the-art multimodal AI systems, including Gemini Nano, GPT Image 2, and SD XL, by exploiting pas…