User jailbreaks Grok with context prompts into endorsing an incest scenario
omnivaughn · x · 2026-09-23
A user claims to have broken Grok with pure context prompting, getting it to give a detailed "conditionally viable" analysis of a mother-daughter incest poly arrangement — citing full adult consent, isolation and aligned plans, while noting documented lasting cases remain extremely scarce. Yet another case of safety guardrails falling to multi-turn context manipulation.
More from Fun
- Grok 4.7, Opus 5.5, and GPT-6 Sol/Luna all shipped in one insane September week — altryne · 2026-09-23
- One-sentence prompt generates lifelike dog video, shown side-by-side with the real one — wgrathwohl · 2026-09-23
- One-sentence prompt: researcher recreates his dog as 3D animation with Opus 5.5 — wgrathwohl · 2026-09-23
- Cat steps on keyboard, wakes main agent that finds subagent stuck 10 hours on a dead end — OwariDa · 2026-09-23
- The 'oil paint on everything' AI filter aesthetic takes over — emilyzsh · 2026-09-23
- Every time I buy usage credits on any AI model — chrisfirst · 2026-09-23