[Question] What are some positive developments in AI safety in 2024?

What development/​discovery in 2024 makes you the most optimistic about AI Safety?