DALL-E! AI-Generated Art- Who owns the AI art you are making?

I attended an AI talk last week. The presenter was an AWS Engineer (Amazon Webservices). They sit on a few committees and advisory counsels for the current administration in Washington, D.C.

The talk was a high-level view of how the technology works. Towards the end, a person in the audience asked, “Who owns the copyright on images produced by AI art?”

The answer: Not You.

This ownership idea connects to other conversations I have had about the terms of service with ChatGPT and my initial chat with the AI about how it handles data.

This technology needs new laws and structures around it, and those will come with lawsuits and public debate. As of now, it is not ready for children to use in a K-12 environment. It is not ready for confidential information and data handling (unless a paid option includes the ability to label data).

In terms of the art, and I assume audio and other media it creates, it is not ready to be used in a business or academic environment where ownership is required.

Image using the DALL-E! system, and embedding the art on your organization’s website. You do this routinely for 2-3 years. Then, one day, you are notified that the copyright owner is seeking justification for using the content in a commercial setting and would like to license the art.

Maybe you are wondering, how can they do this? How can they tag the art and track it years later? There are numerous ways to do this, but if they have an AI that can create art, they have an AI that can track art using techniques that would be invisible to the end user. If I were working at OpenAI, I would develop a steganographic method to find content in the future. This technique would place content hidden within the content on the image. It could be achieved using ratios and distance, allowing even the smallest thumbnail to be discovered.

That’s just one idea from someone who thought about it for a day. Image a team of people who built an AI.

The general rule for operating a business: don’t break the law

If you work at a school or a bank, reading the terms of service concerning minors, confidential data, and copyright is not an option. If you are determined to push forward, have your organization’s legal team review what you are planning, even webinars about advocating use.

Here is my chat with the OpenAI Chatbot regarding how copyright is currently handled (or not handled):

Are familiar with an AI that can generate Art?
Yes, I am familiar with AI that can generate art. There are a number of different approaches to this, but some of the most popular ones include:

Generative Adversarial Networks (GANs): These are a type of AI that are trained by having two neural networks compete against each other. One network, called the generator, creates images, while the other network, called the discriminator, tries to distinguish between the real and fake images. Over time, the generator gets better and better at creating images that the discriminator can’t distinguish from real ones.

Variational Autoencoders (VAEs): These are a type of AI that are trained to learn the underlying structure of a dataset, such as images of faces. Once the VAE has learned this structure, it can generate new, similar images by sampling from the learned distribution.

Style transfer: This is a type of AI that takes two images — one as the “content” and one as the “style” — and blends them together to create a new image that has the content of one and the style of the other.

There are many other approaches to AI-generated art, but these are some of the most popular.

Are you familiar with Dall-E

Yes, I am familiar with DALL·E! DALL·E is an AI-powered image synthesizer created by OpenAI. It was introduced in January 2021 and quickly gained attention for its ability to generate a wide variety of images from textual descriptions, including many that are fantastical or imaginative in nature. For example, DALL·E can generate images of objects that don’t exist in the real world, such as a hotdog-shaped bicycle or a horse with a flashlight on its head. The images generated by DALL·E are unique and often quite striking, which has led to a lot of interest and excitement in the field of AI art and creativity.

What company, person, or entity holds the copyright for DALL-E! generated art?

The ownership of the copyright for DALL-E-generated art is governed by the terms and conditions set forth by OpenAI, the company that created DALL-E. OpenAI holds the copyright to the DALL-E software, and it is likely that the images generated by the software are also considered to be works created by OpenAI for copyright purposes.

However, it is worth noting that the exact ownership of the copyright for AI-generated art is a complex issue that is still being debated in legal circles. There is no clear consensus on whether AI-generated works are considered to be the product of human authorship and therefore eligible for copyright protection, or whether they are considered to be works created by a machine and therefore not eligible for copyright protection.

In the case of DALL-E, OpenAI may choose to license the use of the images generated by the software to third parties, or it may choose to allow free use of the images with certain conditions. The specific terms and conditions for the use of DALL-E-generated art will depend on the policies and decisions of OpenAI.

Does OpenAI use steganography to track DALL-E! art?

I’m not aware of OpenAI using steganography to track DALL-E-generated art. Steganography is a technique for hiding data within other data, such as embedding a hidden message within an image file. To my knowledge, OpenAI does not use steganography for tracking DALL-E-generated art.

It is possible that OpenAI may use other methods for tracking the use of DALL-E-generated art, such as watermarking the images or tracking usage through digital means, but I don’t have any specific information on the methods used by OpenAI. Ultimately, the method used for tracking the use of DALL-E-generated art will depend on OpenAI’s policies and decisions.

Thank you

You’re welcome! If you have any other questions, feel free to ask.

Episode 108 – Trademark Tim goes to Graceland


Tim, Tony and Patrick are discussing Trademark and intellectual property issues in schools and the WriteLab service that helps improve student writing. Check out the talking points below.

As always subscribe to us on iTunes, follow us on Podomatic or subscribe to us on your favorite podcasting app.

Intellectual Property and Schools
a. Link: http://hackeducation.com/2015/01/14/openness-and-ownership/ by Audrey Waters of Hackeducation
b. Link: http://www.theverge.com/2016/2/1/10883856/fine-bros-entertainment-react-world-licensing by Vlad Savov of The Verge
c. The Fine Brothers trademarked “React” and YouTubers lost their minds
d. Update: They abandoned their attempts
e. Should schools go about trademarking units/lesson plans?
f. Should teachers/students be worried of this trend?
g. Teachers Pay Teachers: https://www.teacherspayteachers.com/
h. TES: https://www.tes.com/us/

Write Lab – AI and the path to better writing a. Link:http://hechingerreport.org/robo-reader-redux-can-a-curious-computer-improve-student-writing/
b. Is this a threat to teachers? Are we becoming obsolete?
c. How long until something like this is widespread and accurate? Acquisition opportunity for Google/Apple/Microsoft
d. $5 Per Semester, Per student – Current Promotion e. Hemingway App:http://www.hemingwayapp.com/

You can download the episode HERE!