The 1980's are back in a new way!
Here's a free tip. Guardrails you place to guide user experience will receive circumvention attempts by people with more time, and more creativity than you. This guiding principle is crucial to keep in mind when implementing new tech: No matter how well you think you have designed your guardrails, people will always find ways to circumvent them. Don't believe me?
Check out this informative article on the topic:
ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs
Remember, your guardrails are nothing compared to the creativity of determined users.
News originally broke by Ethan Mollick on Twitter. I'm just a node in the chain: https://x.com/emollick/status/1763687813386547594?s=20
Comments