AI Safety Agendas

More concrete examples of what can be studied during the pause:

• Questions that currently seem most important, Jack Lindsey, 2026
• A transparency and interpretability tech tree, Evan Hubinger, 2022
• A pragmatic vision for interpretability, Neel Nanda et al., 2025
• 200 concrete open problems in mechanistic interpretability, Neel Nanda, 2022
• Open problems in mechanistic interpretability, Lee Sharkey et al., 2025
• Towards developmental interpretability, Jesse Hoogland et al., 2023
• There will be a scientific theory of deep learning, Jamie Simon et al., 2026
• Selective modularity: a research agenda, Alex Cloud and Jacob Goldman-Wetzler, 2025
• Alignment roadmap, AI 2040 supplement

And more here: Shallow Review of Technical AI Safety.