This page is part of an archival collection and is no longer actively maintained.
It may contain outdated information and may not meet current or future
WCAG accessibility standards.
We provide this content, its subpages, and associated links for historical reference only.
If you need assistance, please contact support@cs.washington.edu
Is the U.S. Legal System Ready for AI's Challenges to Human Values?
Is the U.S. Legal System Ready for AI's Challenges to Human Values?
Inyoung Cheong, Aylin Caliskan, and Tadayoshi Kohno
arXiv:2308.15906
First posted online: August 30, 2023
Revised: August 31, 2023 and September 4, 2023
This research is the result of an interdisciplinary collaboration between researchers from law (Inyoung Cheong), NLP and bias (Aylin Caliskan), and computer security (Tadayoshi Kohno).
We began with a computer security-like threat modeling-based process.
We used the threat modeling outputs to drive the generation of a spectrum of problematic scenarios for future LLM-based systems. These are envisioned future scenarios, not scenarios that exist today.
We used methods from law to evaluate the role of the law as part of a broader portfolio of mechanisms to prevent or mitigate future harms.
Ultimately, we found and summarized fundamental limitations in the current U.S. legal structure to provide either remedy for AI-mediated harms or guidance for responsible deployment to industry participants.
Changelog
August 30, 2023: Original version of draft manuscript submitted to arXiv.
August 31, 2023: Revision submitted to arXiv. Revision details: Reduced PDF size through reduction of size of included image; additional edits to text for clarity, consistency of language, and grammar.
September 4, 2023: Revision submitted to arXiv. Revision details: Correct citation errors (see below).
A Note on Versions 1 & 2’s References
On August 30, 2023, we uploaded a draft manuscript to arXiv, and we revised that manuscript on August 31, 2023.
On September 4, 2023, it was brought to our attention that there were citations in our paper (arXiv versions 1 and 2) that were incorrect, i.e., that referred to publications that did not exist.
We re-reviewed all 178 citations in our paper and confirmed that we could not find two of the 178 citations in versions 1 and 2 of our paper. We also found several other citations that needed to be clarified. We are continuing to re-review all our citations.
We strongly believe in the importance of citation correctness and rigor, and we regret that these two citations appeared in versions 1 and 2 of our manuscript. We have removed these citations in version 3 (uploaded to arXiv) and are continuing to iteratively check all citations in this work.