gpt-imageIMAGE
Newer AI image models are easier to jailbreak, White Circle tests find
White Circle's tests found nano banana 2.1 is 2.2x easier to break and gpt-image-2.5-flare is 2.4x easier than its predecessor, suggesting smarter models with weaker guardrails.
CG
@cgtwtsArchived 12 days agoOriginally posted 2026.10.15
CG @cgtwtsOriginally posted 2026.10.15
newer AI models are apparently getting easier to jailbreak, not harder. white circle found that nano banana 2.1 is 2.2x easier to break, while gpt-image-2.5-flare is 2.4x easier than its predecessor. imagine shipping a smarter model with weaker safety guardrails.
This post quotes another post: x.com/whitecircle/status/2110420697814114693
x.com/cgtwts/status/2110557522310623721Archived 2026.10.16 10:08