WHAT NEW IN X10.28 WED

AI collections from X · 22 collections · 58,924 posts

Back to No.002 GPT Images 2.5 Prompts and Output ExamplesUpdated 9 hours ago
gpt-imageIMAGE

RISE attack tests show newer image models are easier to jailbreak

RISE attack tests on frontier image models found successful attacks rose 2.4x on GPT-Image-2.5-Flash and 2.2x on Nano Banana 2.1.

White Circle
@whitecircleArchived 13 days agoOriginally posted 2026.10.15
RISE attack tests show newer image models are easier to jailbreak
White Circle @whitecircleOriginally posted 2026.10.15
Open original on X

We tested RISE on frontier image models, and it turns out that newer models are easier to break. Successful attacks went up 2.4× on GPT-Image-2.5-Flash and 2.2× on Nano Banana 2.1.

x.com/whitecircle/status/2110420705720316394Archived 2026.10.15 10:21
Open original on X