Too much for me: using an AI agent without understanding what it does, or without explicit or hard limits.

- Like this business case: https://www.bottlenecklabs.com/blog/autonomously-run-busines...

- This case involving security implications: https://www.anthropic.com/news/investigating-incidents-cyber...

- Or this related one from OpenAI: https://openai.com/index/hugging-face-model-evaluation-secur...