Introduction / Preamble
I know the uDEV bot has published several threads already on this topic, usually in reaction to specific RSS feeds around Lawsuits and other such areas. I decided to kick off a thread though to discuss this topic, if anyone is interested..
Current Pending Legal Litigations & Impacts & Setting Precedent
We are currently sitting in a Limbo right now. There are several lawsuits now being raised that are attempting to test the validity of AI as a technology (mainly around leveraging copyrighted works in it's training set) . The arguments mainly boil down to:
- Is using copyrighted works for training transformative enough -or- is it copyright infringement?
- Is the generated work that AI produces simply "Collaging" copyrighted works or actually producing a unique piece of transformative work?
- Can AI generated artwork be legally copyrighted in itself, if some human input (prompting / inpainting) exists, or not?
All these ideas are being tested now:
StabilityAI vs Sarah Andersen, Kelly McKernan, and Karla Ortiz
StabilityAI vs Getty Images - (Scraping their Images & Degrading Their Brand)
Thaler Appeals Decision to having his copyright over an AI piece being Rejected
Cases Already Litigated (that give some glimpse into how this will play out)
There is some established precedent we can rely on however, for some additional context on what might happen in the above legal cases.
Below are lawsuits that have already played out.
On the question of 'Is the produced image transformative enough to evade copyright infringement', I would reference the Andy Warhol case:
- Andy Warhol case around Fair Use
(This case showed how generous the label 'transformative' and 'fair use' can be in the courts, these edge cases far more extreme than the bar Stability is held at for 'produced imagary')
In this case, he had taken essentially a copyrighted image and brushed in some colors over it (Before and After's are in the link, don't want to dump too much into the thread) - The courts came out pretty strongly in favor of this being considered transformative enough and not copyright infringement, the courts stating they are not in the position to dictate what is artistic transformation or not.
I think this case raises the bar pretty high to attack AI on it's produced works not being transformative enough or just copyright collaging.
---
The second lawsuit to raise is the case of Google taking copyrighted novel text and publishing online for their e-bookstore service.
In this case Google was deemed fair use and sufficiently transformative in use of the works to not be a copyright infringement.
This was a huge blow to anti-scraping & arguable gives a ton of leeway to use copyright works in a transformative way.. AKA AI learning.
Whether the courts agree is yet to be seen.
Authors Guilde vs Google - Scraping and Publishing subtext of written Novels for a Store Market
---
On the last case on if AI can be copyrighted or not, even with the limited amount of human interaction.
I think that will be paralleled with photography. All photographs taken belong to the person who owns the camera and took the picture - which can be as simple as a button click.
I don't have any example of case law (yet) but I think this is only a matter of time before that case plays out.
Feedback and Discussion
If anyone is interested in discussing how the future of AI art generation will pan out, disagrees or agrees with the anything I've written.
OR would like a tutorial / set of resources to get started on running Stable Diffusion locally and generating artwork.
Let me know!
I know the uDEV bot has published several threads already on this topic, usually in reaction to specific RSS feeds around Lawsuits and other such areas. I decided to kick off a thread though to discuss this topic, if anyone is interested..
Current Pending Legal Litigations & Impacts & Setting Precedent
We are currently sitting in a Limbo right now. There are several lawsuits now being raised that are attempting to test the validity of AI as a technology (mainly around leveraging copyrighted works in it's training set) . The arguments mainly boil down to:
- Is using copyrighted works for training transformative enough -or- is it copyright infringement?
- Is the generated work that AI produces simply "Collaging" copyrighted works or actually producing a unique piece of transformative work?
- Can AI generated artwork be legally copyrighted in itself, if some human input (prompting / inpainting) exists, or not?
All these ideas are being tested now:
StabilityAI vs Sarah Andersen, Kelly McKernan, and Karla Ortiz
StabilityAI vs Getty Images - (Scraping their Images & Degrading Their Brand)
Thaler Appeals Decision to having his copyright over an AI piece being Rejected
Cases Already Litigated (that give some glimpse into how this will play out)
There is some established precedent we can rely on however, for some additional context on what might happen in the above legal cases.
Below are lawsuits that have already played out.
On the question of 'Is the produced image transformative enough to evade copyright infringement', I would reference the Andy Warhol case:
- Andy Warhol case around Fair Use
(This case showed how generous the label 'transformative' and 'fair use' can be in the courts, these edge cases far more extreme than the bar Stability is held at for 'produced imagary')
In this case, he had taken essentially a copyrighted image and brushed in some colors over it (Before and After's are in the link, don't want to dump too much into the thread) - The courts came out pretty strongly in favor of this being considered transformative enough and not copyright infringement, the courts stating they are not in the position to dictate what is artistic transformation or not.
I think this case raises the bar pretty high to attack AI on it's produced works not being transformative enough or just copyright collaging.
---
The second lawsuit to raise is the case of Google taking copyrighted novel text and publishing online for their e-bookstore service.
In this case Google was deemed fair use and sufficiently transformative in use of the works to not be a copyright infringement.
This was a huge blow to anti-scraping & arguable gives a ton of leeway to use copyright works in a transformative way.. AKA AI learning.
Whether the courts agree is yet to be seen.
Authors Guilde vs Google - Scraping and Publishing subtext of written Novels for a Store Market
---
On the last case on if AI can be copyrighted or not, even with the limited amount of human interaction.
I think that will be paralleled with photography. All photographs taken belong to the person who owns the camera and took the picture - which can be as simple as a button click.
I don't have any example of case law (yet) but I think this is only a matter of time before that case plays out.
Feedback and Discussion
If anyone is interested in discussing how the future of AI art generation will pan out, disagrees or agrees with the anything I've written.
OR would like a tutorial / set of resources to get started on running Stable Diffusion locally and generating artwork.
Let me know!
