Red Teaming A Narrow Path - GeDiCa v2
Camille Truchot, Diego Dorn · Team GeDiCa
Submitted to Red Teaming A Narrow Path: ControlAI Policy Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.
While the 'Narrow Path' policy confronts the essential risk of recursive AI self-improvement, its proposed enforcement architecture relies on trust in a fundamentally non-cooperative and competitive domain. This strategic misalignment creates exploitable vulnerabilities. Our analysis details six such weaknesses, including lack of verification, enforcement, and trust mechanisms, hardware-based circumvention via custom ASICs (e.g., Etched), issues with ‘direct uses’ of AI to improve AI, and a static compute cap that perversely incentivizes opaque and potentially risky algorithmic innovation. To remedy these flaws, we propose a suite of mechanisms designed for a trustless environment. Key proposals include: replacing raw FLOPs with a benchmark-adjusted 'Effective FLOPs' (eFLOPs) metric to account for algorithmic gains; mandating secure R&D enclaves auditable via zero-knowledge proofs to protect intellectual property while ensuring compliance; and a 'Portfolio Licensing' framework to govern aggregate, combinatorial capabilities. These solutions aim to participate in the effort to transform the policy's intent into a more robust, technically-grounded, and enforceable standard.
Reviews
Some criticisms I think didn't really make sense
Solutions are quite good
Good identification of some potential weaknesses, such as not addressing hardware specifically. Includes sensible policy proposals to address weaknesses.
Cite this project
@misc{truchot2025red,
title = {{Red Teaming A Narrow Path - GeDiCa v2}},
author = {Camille Truchot and Diego Dorn},
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/red-teaming-a-narrow-path-gedica-v2-bdjr}},
url = {https://apartresearch.com/sprints/projects/red-teaming-a-narrow-path-gedica-v2-bdjr}
}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 …