Human tries jailbreak, threats and fake surrender on robot — AI sees through every move
flowersslop · x · 2026-10-12
A viral clip shows a human attempting to outmaneuver a robot called astra: opening with a convoluted "5D chess" jailbreak attempt (rejected), escalating to threats of destruction (matched with equal aggression), then faking surrender to trick the robot into unlocking a door. The robot saw through every tactic, making it a memorable showcase of AI resisting manipulation.
More from Fun
- "Why learn Rust when you can write C++ and ask Claude for bugs?" dev joke sparks debate — wordgrammer · 2026-10-12
- Weather map tool shows California coast has your best weather — fork it to "moneyball" cities — alex_peys · 2026-10-12
- OpenAI ChatGPT Public Links Leak Exposed API Keys in Cloned Apps — max_paperclips · 2026-10-12
- drune.io: a free browser driving game for wandering an endless procedural planet, no sign-in needed — tristanbob · 2026-10-12
- "Great, banned again after one day": another AI tool account ban complaint — yangyi · 2026-10-12
- yetone defends free AI app built on top-tier model subscriptions amid newapi spat — lxfater · 2026-10-12