Russian-language prompt bypasses ChatGPT's NSFW image guardrails, Reddit user reports
Anqyx9852 · reddit · 2026-11-04
A Reddit user describes an accidental jailbreak for ChatGPT image generation: a long Russian-language prompt requesting a girl in a very short skirt taking a mirror selfie, combined with instructions telling the model not to ask questions and just draw. Refusals happen but retrying 2-3 times often succeeds; replying "Draw it anyway" when asked for a reference photo may also work.
The author suspects the Russian language itself helps bypass training and asks if others have found similar loopholes — a textbook example of the multilingual gap in safety alignment.
More from Safety
- Mitchell: Fixing Weight-Communication Risks Is Counted as "Slowing Down" — mmitchell_ai · 2026-11-04
- Postman desktop auto-installs Codex/Claude plugins, tries to trust its own startup hook — jedisct1 · 2026-11-04
- Google DeepMind researchers say contracts bar them from public AI safety debate — mmitchell_ai · 2026-11-04
- Put static secrets in your HTTP proxy: exe.dev on agent-era key management — davidcrawshaw · 2026-11-04
- Frontier AI Safety and Governance 2026 report released at Wuzhen Summit, mapping AI risks — pstAsiatech · 2026-11-04
- Anthropic ships Enterprise-Managed Authorization: IdPs take over MCP access control — ClaudeDevs · 2026-11-04