What are some open research questions in AI alignment?

https://www.alignmentforum.org/posts/5HtDzRAk7ePWsiL2L/open-problems-in-ai-x-risk-pais-5

AI alignment has numerous subdomains and each has many open research questions

Here are some examples:

Agent foundations

Mechanistic interpretability - This is research which aims to understand the inside workings of Neural networks

brain-like AI safety

Honesty

Governance

As a new field, there are many more open problems as well, and new questions are constantly being asked