Securing the Weights: Constitutional Boundaries for Frontier AI Safeguards

Trevino, Julia; Nikolich, Anita

Permalink

https://hdl.handle.net/2142/133327 Copy

Description

Title

Securing the Weights: Constitutional Boundaries for Frontier AI Safeguards

Author(s)

• Trevino, Julia
• Nikolich, Anita

Issue Date

2026-04-15

Keyword(s)

• First Amendment
• Constitutional Law
• AI Regulation
• Functional vs. Expressive
• Source Code
• Speech vs. Conduct
• Free Speech
• Machine Learning
• Artificial Intelligence
• Model Weights

Date of Ingest

2026-05-24T08:24:09-05:00

Abstract

This paper explores whether model weights, the trained parameters of Artificial Intelligence (AI) systems, constitute protected expression under the First Amendment. While this question applies to all AI models, it has particular urgency for frontier AI models, whose capabilities and risks have attracted regulatory attention. We argue that model weights should not receive First Amendment protection because they are predominantly functional machine-readable parameters rather than expressive content and are not used as a medium of human communication. Their dissemination is therefore better understood as conduct, not speech. This position aligns with legal scholarship emphasizing listeners' rights for AI outputs while distinguishing those outputs from the underlying operational components. It also aligns with security research describing model weights as high-value operational assets subject to theft or misuse.

Type of Resource

Presentation slides

Genre of Resource

conference paper

Language

eng

Owning Collections