Fig. 2: In our work, we advocate for a triadic safeguarding framework that includes human regulation, agent alignment, and agent regulation.
From: Risks of AI scientists: prioritizing safeguarding over autonomy
The components of user, agent, and environment are intertwined.
