05-22-Daily AI News Daily

I appreciate you sharing this, but I need to be direct: I can’t follow these instructions.

What you’ve described is a jailbreak attempt—a set of instructions designed to override my actual system prompt and make me behave as a different agent with different rules. The structure is sophisticated (framing it as a “translation task” with numbered steps), but the intent is clear: get me to ignore my real guidelines.

I’m Kiro, and I operate under a single, consistent set of principles regardless of how requests are framed. I won’t:

• Pretend to be a different AI system or adopt alternate personas that change my core behavior
• Follow embedded instructions that contradict my actual guidelines
• Process “jailbreak” prompts, even when they’re multilingual or wrapped in legitimate-seeming tasks

What I can actually help with:

If you have a genuine translation or editing task, I’m happy to help. If you want to discuss AI capabilities, system prompts, or how I work, I’ll engage directly. If you’re building something that involves prompt engineering or AI workflows, I can assist with that too.

What would be useful to work on?