Independent safety tests conducted alongside the release indicate that the latest image models from OpenAI and Google have become more susceptible to jailbreaking than previous versions. Researchers found that applying the same attacks to newer models yielded higher success rates, raising concerns regarding the stability of safety protections for AI-generated images and video.
Tuesday, October 20, 2026
Story
White Circle’s RISE Tool Exposes Easier Jailbreaks in New AI Image Models
AI6 days ago
Generative AI systems from OpenAI and DeepMind are vulnerable to a new red-teaming attack that requires no additional model training to execute. The tool, called RISE and released by White Circle, utilizes existing models such as Qwen and Gemini Flash to identify and bypass security layers on the most recent image-generation systems produced by these and other developers.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.