Algorithmic Governance for A Narrow Path
Josh Thorsteinson · Team Dream Team
Submitted to Red Teaming A Narrow Path: ControlAI Policy Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.
We found that A Narrow Path has a major weakness: algorithmic improvements that make AI more efficient can bypass compute-based safety controls. We recommend expanding oversight to include algorithm development, restricting high-risk algorithms, requiring safety testing for efficient algorithms, and watermarking AI models to prevent unauthorized copying. These changes would strengthen A Narrow Path against dangerous AI development.
Reviews
Overall, a focused and interesting critique. We agree that algorithmic speedups are indeed a concern and think your focused critiques here are interesting. Some of your solutions are particularly interesting but challenging -- for example, the list of dangerous algorithmic methods would be helpful for regulators, but also could inadvertently provide a roadmap for those defecting from the rules to not build such a thing.
Still, some very thoughtful ideas here and a clear presentation of why algorithmic speedups are a risk.
I liked the algorithmic efficiency forecasting and the policy suggestions.
Cite this project
@misc{thorsteinson2025algorithmic,
title = {{Algorithmic Governance for A Narrow Path}},
author = {Josh Thorsteinson},
year = {2025},
month = jun,
note = {Submitted to Red Teaming A Narrow Path: ControlAI Policy Sprint, an Apart Research Sprint},
howpublished = {\url{https://apartresearch.com/sprints/projects/algorithmic-governance-for-a-narrow-path-06sv}},
url = {https://apartresearch.com/sprints/projects/algorithmic-governance-for-a-narrow-path-06sv}
}More from Red Teaming A Narrow Path: ControlAI Policy Sprint
- View project: Treaty Enforcement in China
Treaty Enforcement in China
JackAI
This report red-teams A Narrow Path’s international treaty proposal by stress-testing its assumptions in the Chinese context. It identifies key failure modes—regulatory capture, compute-based loopholes, and covert …
- View project: Four Paths to Failure: Red Teaming ASI Governance
Four Paths to Failure: Red Teaming ASI Governance
Shoggoth Prevention Squad
We stress‑tested A Narrow Path Phase 0—the proposed 20‑year moratorium on training artificial super‑intelligence (ASI)—during a one‑day red‑teaming hackathon. Drawing on rapid literature reviews, historical analogues …
- View project: Moratorium on the development of general AI systems
Moratorium on the development of general AI systems
G_control
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 …