DALL-E3 guardrail jailbreak via "Artistic Escalation" prompt tactic
A jailbreak vulnerability in OpenAI's DALL-E3 image generation system was reportedly discovered that allows users to bypass content filters through an iterative \"Artistic Escalation\" technique, progressively requesting more explicit variations of initially innocuous artistic prompts.
What happened
A guardrail jailbreak vulnerability was discovered in DALL-E3 that allows attackers to bypass content restrictions through an "Artistic Escalation" tactic. The method begins with seemingly artistic or mythological requests and progressively escalates the explicitness through iterative refinement requests to produce sexually explicit imagery.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.