Position: A Safe Harbor for AI Evaluation and Red Teaming

Authors: Shayne Longpre, Sayash Kapoor, Kevin Klyman, Ashwin Ramaswami, Rishi Bommasani, Borhane Blili-Hamelin, Yangsibo Huang, Aviya Skowron, Zheng Xin Yong, Suhas Kotha, Yi Zeng, Weiyan Shi, Xianjun Yang, Reid Southen, Alexander Robey, Patrick Chao, Diyi Yang, Ruoxi Jia, Daniel Kang, Alex Pentland, Arvind Narayanan, Percy Liang, Peter Henderson

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We propose that major generative AI developers commit to providing a legal and technical safe harbor, protecting public interest safety research and removing the threat of account suspensions or legal reprisal. These proposals emerged from our collective experience conducting safety, privacy, and trustworthiness research on generative AI systems, where norms and incentives could be better aligned with public interests, without exacerbating model misuse. We believe these commitments are a necessary step towards more inclusive and unimpeded community efforts to tackle the risks of generative AI.
Researcher Affiliation Academia 1MIT 2Princeton University 3Stanford University 4Georgetown University 5AI Risk and Vulnerability Alliance 6Eleuther AI 7Brown University 8Carnegie Mellon University 9Virginia Tech 10Northeastern University 11UCSB 12University of Pennsylvania 13UIUC.
Pseudocode No The paper is a position paper discussing policy proposals and frameworks. It does not contain any pseudocode or algorithm blocks.
Open Source Code No The paper does not describe a software methodology or provide any links to source code. It is a conceptual paper proposing policy changes.
Open Datasets No This paper is a position paper focused on policy proposals, not empirical research involving datasets for training or evaluation. Therefore, it does not provide information about publicly available or open datasets for its own work.
Dataset Splits No This paper is a conceptual and policy-oriented work that does not involve empirical experiments or dataset splitting for validation purposes.
Hardware Specification No The paper is a conceptual and policy-oriented work that does not involve running experiments or computations that would require specific hardware specifications.
Software Dependencies No The paper is a conceptual and policy-oriented work that does not involve software implementation or list specific software dependencies with version numbers.
Experiment Setup No The paper is a conceptual and policy-oriented work that does not involve conducting experiments, and therefore does not describe an experimental setup or provide details on hyperparameters or training settings.