# Can an AI check its own safety?

**Can an AI check that its successor is safe?**

A system that wants to trust a stronger version of itself meets a limit like Löb’s. Humans and other checks may be part of the answer.

Connects to:
- [Making things safe](https://mapideas.org/i/te_safety)
- [the AI version](https://mapideas.org/i/h_ai)
- [Löb’s theorem](https://mapideas.org/i/h_lob)

Source: https://mapideas.org/i/q_x_ai_check · open on the map: https://mapideas.org/#q_x_ai_check