Abstract
This disclosure describes a system that can adapt code signing principles to token sequences used as input for LLMs (e.g. natural language) by, for example, treating LLM inputs as signed artifacts. An artifact can be a structured data object containing the token sequence, metadata, and a digital signature created with a private key. An integrated policy enforcement point may intercept incoming sequence, cryptographically verify the artifact's signature against a configurable trust store of public keys, and enforce access policies before the content enters the LLM's context. This system can help establish verifiable provenance, safety and integrity for external data, creating an auditable, cryptographically-secured boundary for the LLM.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Tretyakov, Konstantin, "System for Cryptographic Signing and Verification of Textual Inputs for LLM Agents", Technical Disclosure Commons, (April 16, 2026)
https://www.tdcommons.org/dpubs_series/9803