Breaking AI With... Poetry?

Thanks to our LMG clips sponsors dbrand, Dell, and Secretlab. You can check them out at the links below: dbrand: https://dbrand.com/pcb Dell: https://lmg.gg/dellprowanclips Secretlab: https://lmg.gg/secretlablmgclips Linus and Luke discuss a new study showing that “adversarial poetry” can reliably bypass AI safety guardrails. They break down how major models responded, why creative writing can outperform technical exploits, and what this means for the future of AI safety. Watch the full WAN Show: https://www.youtube.com/watch?v=35igp1OBRY0 ► GET MERCH: https://lttstore.com ► GET EXCLUSIVE CONTENT ON FLOATPLANE: https://lmg.gg/lttfloatplane ► SPONSORS, AFFILIATES, AND PARTNERS: https://lmg.gg/partners