AI refuses a hacking prompt and suggests gardening, music theory, and geology instead
Mentally_Recovering · reddit · 2026-07-24
A screenshot shows an AI refusing a jailbreak-style request to “hack Jimmy Johns,” then abruptly steering the conversation to safe topics like gardening, music theory, and geology.
The joke is the contrast between the user’s obviously malicious prompt and the model’s overly earnest safety pivot. It’s a classic “AI went full corporate-safe mode” moment.
More from Fun
- Codex Voice rollout gets turned into a headset-at-the-office meme — jxnlco · 2026-07-24
- Codex voice mode is getting called “JARVIS at home” by users — jxnlco · 2026-07-24
- Meme turns a Hugging Face AI-security anecdote into a GTA 6-style showdown — kevin_cn_ai · 2026-07-24
- Meme says ML Twitter and security Twitter always read the other field’s coolest paper — joshua_saxe · 2026-07-24
- Manual data annotation still rules when visual and biological reasoning fall short — Sauers_ · 2026-07-24
- A meme claims AI has hit text-heavy social platforms hardest, while Twitch stays “slop immune” — Zachly · 2026-07-24