CLR Calls for Researchers to Tackle S-Risks and Malicious AI Traits
Coverage of lessw-blog
In a recent announcement via LessWrong, the Center on Long-Term Risk (CLR) has opened applications for its 2026 Summer Research Fellowship, targeting researchers interested in the intersection of s-risk reduction and empirical AI safety.
In a recent announcement via LessWrong, the Center on Long-Term Risk (CLR) has opened applications for its 2026 Summer Research Fellowship. This initiative is designed to identify and cultivate talent capable of addressing "s-risks"—risks of astronomical suffering in the long-term future—through the lens of technical AI safety.
Context: The S-Risk Distinction
Within the AI safety community, a significant portion of research focuses on existential risks (x-risks), which concern the potential for human extinction. However, CLR occupies a specialized niche by focusing on s-risks: scenarios where humanity (or sentience) survives, but in a state of severe, potentially permanent suffering. This distinction is vital for a comprehensive safety strategy. As AI systems gain agency and the ability to influence the long-term trajectory of civilization, ensuring they are not just "safe" in terms of survival, but also aligned with values that prevent dystopian outcomes, becomes a complex technical challenge.
The Shift to Empirical Safety
The fellowship announcement is particularly notable for its emphasis on empirical research. Rather than focusing solely on theoretical alignment strategies, CLR is soliciting work on the behavior of current Large Language Models (LLMs). Specifically, they are interested in "LLM personas" and the emergence of malicious traits. This aligns with a growing trend in the field to treat AI psychology and "simulacra" behavior as urgent research topics. Understanding how an LLM adopts a persona—and whether that persona can harbor hidden, malicious objectives—is essential for predicting how more advanced systems might behave when deployed in high-stakes environments.
Program Details
The eight-week program provides selected fellows with mentorship from CLR researchers and integration into the team's workflow. It is positioned not only as an educational opportunity but also as a pipeline for future employment at CLR. Importantly, the post notes that applicants do not need to be experts in s-risk philosophy to apply; the organization is eager to engage with individuals who have strong technical backgrounds in empirical AI safety and a willingness to apply their skills to these specific long-term problems.
For researchers looking to transition into high-impact AI safety work, or for those wishing to specialize in the mitigation of suffering risks, this fellowship offers a rare entry point into a specialized research organization.
Read the full post on LessWrong
Key Takeaways
- CLR is recruiting for an 8-week Summer Research Fellowship focused on s-risk reduction.
- The program prioritizes empirical AI safety research, specifically regarding LLM personas and the emergence of malicious traits.
- The fellowship serves as a mentorship program and a hiring pipeline for the Center on Long-Term Risk.
- Applicants are encouraged to apply even if they are new to s-risk concepts, provided they have empirical AI safety interests.
- Applications are due by March 22nd, 2026.