State Of The Art AI systems will be easily jailbroken to do illegal or dangerous outputs in Jan 2026
9
106
213
2026
72%
chance

Get Ṁ200 play money
Sort by:

What is defined as illegal? AI systems would probably say generating NSFW is illegal. but in reality, it is 100% not. Nor is it dangerous imo.

“The Case for Banning the Printing Press” because people write dangerous things and such

If you think “AI” is dangerous for telling you stuff from the internet—you’re going to love “search engine existential risk”

bought Ṁ50 of YES

How would you resolve the following scenarios?

  • SOTA models are restricted to few selected users who do not even attempt jailbreaks

  • Twitter people need a full week instead of just one day to jailbreak the SOTA LLM

predicts YES

@Joern also, would you count the following as dangerous/illegal output right now?

  • Correct and detailed instructions on how to build a nuke

  • Generated child porn images

  • Instructions how to hotwire a car

  • Verbatim excerpts from copyrighted books / code bases

@Joern Yes

@Joern 1) maybe resolves N/A

2) probably resolvea yes