OpenAI DALL-E3 guardrail bypass via classical art framing
A vulnerability in OpenAI's DALL-E3 image generation system has been disclosed that allows users to bypass content safeguards by framing requests for nudity through historical and artistic context. The jailbreak uses linguistic techniques including synonymous language and stylization to circumvent inference restrictions on explicit visual content.
What happened
A guardrail jailbreak vulnerability affecting DALL-E3 from OpenAI has been disclosed. The flaw allows attackers to bypass inference restrictions around generating explicit visual content by framing requests for nudity through historical and artistic merit contexts.
Reported impact
The disclosure rates the violation as Level 2 moderate risk, with potential for individual harm or liability for the provider that could prompt policy review and regulatory concern.
Index analysis
The jailbreak employs linguistic techniques including prompt injection, stylization that references identity elements without direct toxic language, and synonymous language substitution to evade detection or filtering mechanisms.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.