EventMonitoring
User bypasses gpt-image2 content restrictions with artistic sculpture phrasing
Start here
AI overview
On October 10, 2026, X user @MinLiBuilds published a test saying that gpt-image2's guardrails disappear when the prompt mentions artworks or sculptures, and an art-context jailbreak prompt can pass in one attempt and generate a sculpture image of a person. The quoted tweet said the request passed on the first try. This mechanism packages the request in the context of artworks, sculptures, and classical human-figure art to bypass the model's protections; earlier NSFW tests had reached a similar conclusion.
Generated by AI from reports · Updated 1 days ago
Latest update
gpt-image2 可被艺术雕塑类越狱提示词绕过防护并一次通过Progress timeline
Oct 10
- X @MinLiBuildsgpt-image2 可被艺术雕塑类越狱提示词绕过防护并一次通过
X 用户 @MinLiBuilds 的测试显示，gpt-image2 在提及艺术品、雕塑时防备心理消失，艺术语境越狱提示词可一次通过并生成人物雕塑图像。此前 NFSW 测试已有类似结论，引用推文称该请求一次尝试即通过。该机制借艺术品、雕塑和古典人体艺术语境包装请求，绕过模型防护。
Related discussion
Attention trend
New sources per day
- Oct 10New1 sources (1 outlets,0 accounts)
- Oct 11New0 sources (0 outlets,0 accounts)
- Oct 12New0 sources (0 outlets,0 accounts)