AI Use Restriction Addendum, version 1.0

README.md

AI Use Restriction Addendum, version 1.0

This gist contains a universal AI Use Restriction Addendum that can be attached to many open source licenses in order to restrict use of the licensed work for training AI systems, or as a component in a larger derivative work that uses or integrates with AI systems, either locally (e.g. via libraries, SDKs, etc.) or via remote APIs.

It was created with the assistance of a senior IP lawyer licensed in the United States, and builds upon existing work, notably EU AI Act Article 3(1).

Applying the Addendum to Your Own Works

To use the addendum, append the text contained in AI-Use-Restriction-Addendum-1.0.txt to the existing open source license text in your LICENSE and/or COPYING file(s).

If you're using an SPDX identifier to communicate how your work is licensed (perhaps in a build tool or package manager metadata file), you should also append the expression WITH AdditionRef-AI-Use-Restriction-Addendum-1.0 to that existing SPDX identifier, thereby creating a composite SPDX license expression. For example, a project that was previously licensed under MIT would become MIT WITH AdditionRef-AI-Use-Restriction-Addendum-1.0.

License

The addendum itself is copyright (c) Peter Monks 2026, and distributed under the CC0-1.0 license.

FAQ

Q. Is the resulting license + addendum still "open source"?
A. In practice it will be as open source as the parent license you're already using, for downstream users not engaged in training or constructing AI systems or services. But according to the OSI's Open Source Definition, adding this addendum to an open source license would violate criteria 6 (by discriminating against AI systems and services) and make the resulting work not open source.

Whether the OSI's definition of "open source" is relevant or not is entirely up to you to decide, in the context of your own work. This addendum is intended to be more philosophically aligned with the blossoming Ethical Source movement rather than the legacy Open Source movement.

Q. Aren't AI companies using the fair use doctrine to do an end run around copyright and licensing laws?
A. Yes they are, with mixed success in the courts to date.

Q. Doesn't fair use make this addendum meaningless?
A. Not necessarily. Clearly stating (via license terms such as these) that you don't want your IP used in these systems is better than doing nothing.

AI-Use-Restriction-Addendum-1.0.txt

AI‑USE RESTRICTION ADDENDUM, Version 1.0

This Addendum imposes additional conditions restricting certain uses of the
Licensed Work in connection with Artificial Intelligence systems. When properly
attached or referenced by the Licensor, this Addendum forms part of the license
terms governing the Licensed Work.

0. DEFINITIONS

0.1 "Licensed Work" means the software, source code, object code, documentation,
data, and other copyrightable materials to which this Addendum is attached or
with which it is distributed, and any portion thereof.

0.2 "AI System" means:

(a) a machine‑based system designed to operate with varying levels of autonomy
and that may exhibit adaptiveness after deployment, and that, for explicit or
implicit objectives, infers, from the input it receives, how to generate outputs
such as predictions, content, recommendations, or decisions that can influence
physical or virtual environments; and

(b) any system, model, or component commonly described as artificial
intelligence, machine learning, deep learning, neural networks, large language
models (LLMs), foundation models, multimodal models, or general‑purpose AI
models, whether accessed locally or remotely, including as weights, parameters,
embeddings, adapters, model artifacts, hosted services, APIs, SDKs, or
endpoints.

0.3 "AI Service" means any hosted, cloud, remote, or third‑party service
(including via API, SDK, or endpoint) that provides access to or functionality
of an AI System, including inference, content generation, ranking,
recommendations, classification, clustering, detection, scoring, or decision
support.

0.4 "AI‑Enabled Product or Service" means any software, product, system, or
service—whether external‑facing or internal‑only—that:

(a) uses, invokes, calls, integrates with, depends upon, or is designed to
operate in conjunction with any AI System or AI Service; or

(b) provides, enables, or materially supports AI‑based functionality (including
generation of text, code, images, audio, video, predictions, recommendations,
classifications, or decisions), regardless of whether the AI System is embedded
locally, linked, bundled, containerized, or accessed remotely through an AI
Service.

0.5 "Train" or "Training" means any use of the Licensed Work (or any Derivative
Works) as input, in whole or in part, to develop, improve, tune, calibrate,
validate, test, benchmark, evaluate, or otherwise adjust or assess an AI System
or any component thereof, including without limitation: pre‑training, fine‑
tuning, instruction tuning, reinforcement learning (including RLHF),
distillation, embedding/representation learning, gradient updates, parameter
estimation, model selection, hyperparameter optimization, synthetic data
generation for training, or dataset curation performed for any of the foregoing
purposes.

0.6 "Derivative Works" means any work based on or derived from the Licensed Work
under applicable copyright law, including modifications, translations,
adaptations, or combinations containing non‑trivial portions of the Licensed
Work.

0.7 "Use" includes (without limitation) to access, run, execute, process, copy,
reproduce, distribute, transmit, host, publish, display, perform, modify, create
derivative works of, integrate, link, compile, bundle, embed, or otherwise
exploit the Licensed Work, whether directly or indirectly, and whether for
internal purposes, for third parties, or for external deployment.

1. PROHIBITED USES (CONDITIONS OF LICENSE)

1.1 NO TRAINING. You must not Train, and must not permit any third party to
Train, any AI System using the Licensed Work or any Derivative Works, in whole
or in part.

1.2 NO AI‑ENABLED USE (INCLUDING INTERNAL USE AND API USE). You must not Use,
and must not permit any third party to Use, the Licensed Work or any Derivative
Works as part of, in connection with, or for the benefit of any AI‑Enabled
Product or Service. This prohibition includes, without limitation:

(a) incorporating, linking, compiling, bundling, embedding, distributing,
deploying, or hosting the Licensed Work within an AI‑Enabled Product or Service;

(b) using the Licensed Work in any product or service (including internal‑only
tools) that calls, invokes, or otherwise uses an AI Service (including any LLM
API), whether or not the AI System is hosted by You or a third party; and

(c) using the Licensed Work to enable, support, enhance, or provide
functionality to an AI System or AI Service, including middleware, plugins,
connectors, adapters, prompt orchestration, retrieval‑augmented generation (RAG)
components, agent frameworks, classifiers, recommenders, ranking systems, or
inference pipelines.

1.3 NO CIRCUMVENTION / NOTICE REMOVAL. You must not remove, obscure, or alter
licensing, provenance, or rights‑reservation notices intended to communicate the
restrictions of this Addendum.

2. TERMINATION

2.1 AUTOMATIC TERMINATION. Any violation of Section 1 immediately and
automatically terminates any license granted to You in the Licensed Work (in
addition to any other remedies available to the Licensor).

3. DOWNSTREAM NOTICE

3.1 REDISTRIBUTION. Any redistribution of the Licensed Work or Derivative Works
must include this Addendum or a URL to the canonical text of this Addendum, in a
manner reasonably calculated to provide notice to recipients.

END OF ADDENDUM

COPYING

Copyright (c) Peter Monks 2026
Distributed under the CC0-1.0 license - see https://creativecommons.org/publicdomain/zero/1.0/legalcode.txt