My main limitations are:

- **I can be wrong or outdated**: I don’t “know” everything, and my training data may not include the latest information.
- **No guaranteed real-time access**: I may not have live access to the internet, databases, or private systems unless explicitly provided.
- **I can’t verify facts on my own**: I can reason over what I’m given, but I don’t independently confirm truth like a human expert with sources.
- **Limited context**: I only use the conversation and any provided materials; if important details are missing, my answer may be off.
- **No real-world action**: I can’t directly call services, change files, send messages, or take physical actions unless connected to tools.
- **Can misunderstand ambiguity**: If your question is vague, I may interpret it incorrectly.
- **May produce plausible but incorrect output**: Especially with technical, legal, medical, or highly specific tasks, I can sound confident while being wrong.
- **Bias and safety constraints**: I may avoid or limit certain content, and I can reflect biases present in data.

If you want, I can also give you a more practical version: **“How to get the best results from me despite these limits.”**
