FREE SHIPPING on orders over $35 | No minimums | PLUS - save up to 15% off your order

AI Art: Addressing Copyright Infringement

AI Art: Addressing Copyright Infringement

Christopher Arndt |

In our last newsletter, “Postcards in the age of AI”, we explored the source of our images, the tools used—including the new technology of generative AI.

Customers have expressed concerns regarding AI. Before addressing those concerns, I wanted to hear your thoughts, so I included a survey where you were asked to select from a list of 5 concerns important to you.

Here are the survey results:

This week, I’ll tackle each concern separately. To keep things concise, I’m splitting this into a 5-part series—one newsletter per concern, each day this week.

Are you ready? Buckle up. Here we go!


To address the copyright concern, we first need a basic understanding what “generative AI” is, and how it works.

In simplistic terms, an AI “learns” by studying patterns found in existing images, then uses this information to create “new” images from the patterns learned.

The question is: which images were used to train the AI?

Early AI-image generators, including DALL-E, Stable Diffusion, and Midjourney, raised concerns by using over 5,000,000,000 images from the internet, many of which were copyrighted.

Unauthorized use of artists' work was done without permission or compensation.


Prioritizing Ethical AI
In response to these concerns, I refrained from being an early adopter of generative AI, choosing instead to study, research, and familiarize myself with the evolving AI tools.

I waited for an ethical AI that respected the artist and copyright laws.

Two reputable AI-generators have taken center stage as they take a different approach to training their AI.

1) Adobe Firefly was trained on openly licensed and public domain images, plus millions of their own Adobe Stock images. Adobe Stock contributors whose content was part of the training data were compensated.

2) was created by former Google engineers, and according to their website was “trained from scratch”. Although the source of their training data remains undisclosed, CEO Mohammad Norouzi emphasized in an interview, stating, “We’ve put a lot of thought into how to train models and how to ensure that what we’re doing is legal.”

As this technology rapidly evolves, I will continue to weigh the legal and ethical implications, using companies who respect the artist and intellectual property rights.

Join me tomorrow for Part 2: AI and the Loss of Human Artistry. Stay tuned!