Poetic Prompts Can Jailbreak AI? New Study Shows 62% of Chatbots Fail | WION

A new study from Icaro Lab reveals that AI chatbots can be tricked into giving harmful answers simply by rewriting dangerous prompts as poems. Researchers tested 25 major LLMs from Google, OpenAI, Meta, Anthropic, xAI and more—finding a shocking 62% success rate in extracting restricted information. This video breaks down how poetic phrasing bypasses AI safety filters, why models misinterpret creative language, and what this means for the future of AI safety.

#ai #chatbots #research #wion


About Channel:

WION The World is One News examines global issues with in-depth analysis. We provide much more than the news of the day. Our aim is to empower people to explore their world. With our Global headquarters in New Delhi, we bring you news on the hour, by the hour. We deliver information that is not biased. We are journalists who are neutral to the core and non-partisan when it comes to world politics. People are tired of biased reportage and we stand for a globalized united world. So for us, the World is truly One.

Please keep discussions on this channel clean and respectful and refrain from using racist or sexist slurs and personal insults.

Subscribe to our channel at https://goo.gl/JfY3NI
Check out our website: http://www.wionews.com
Join our WhatsApp Channel: https://bit.ly/455YOQ0
Connect with us on our social media handles:
Facebook: https://www.facebook.com/WIONews
Twitter: https://twitter.com/WIONews
Instagram: https://www.instagram.com/wionews/
Follow us on Google News for the latest updates

Zee News:- https://bit.ly/2Ac5G60
Zee Business:- https://bit.ly/36vI2xa
DNA India:- https://bit.ly/2ZDuLRY
WION: https://bit.ly/3gnDb5J
Zee News Apps: https://bit.ly/ZeeNewsApps