Jun 14, 2025
Red Teaming A Narrow Path: ControlAI Policy Sprint
Gideon Daniel Giftson T
All six policies are red teamed step-by-step systematically. We initially corrected vague definitions and also found that
the policies regarding the capabilities of AI systems lack technical soundness and that more incentives are needed to entice states to sign the treaty. Further, we discover a lack of equity in the licensing framework, and a lack of planning for black-swan events. We propose an oversight framework right from the manufacturing process of silicon chips. We also propose calling for a moratorium on the development of general AI systems until the existing tools for analyzing them can catch up. Following these recommendations still won't guarantee the prevention of ASI for 20 years, but ensures that the world is on track to even tackle such a system if it is somehow created.
Tolga Bilge
Thoughtful analysis, points out a number of potential issues such as sandbagging, and the lack of incentives to join the treaty system as formulated in Phase 0
Cite this work
@misc {
title={
@misc {
},
author={
Gideon Daniel Giftson T
},
date={
6/14/25
},
organization={Apart Research},
note={Research submission to the research sprint hosted by Apart.},
howpublished={https://apartresearch.com}
}