A site is stealing and duplicating our GPTs - how can we protect our GPTs?

GPT builders
, ,
21

I find it unlikely that OpenAI would dispense free advice in protecting your intellectual property.

Let’s discuss IP mechanisms:

Trademark: government-issued work mark or design protection that provides you a course for remedy in the case of dilution of trade by impostors in your area of business - once granted.

(FYI: OpenAI had both “GPT” and “ChatGPT” denied trademark so far by the USPTO. That’s likely why you see a new attempt at propagation and dilution of the value of the name as meaning “generative pretrained transformer”.)

Copyright: In the US, is automatically granted on your creative works that are not solely fact-based. Works can still be registered for more notation by the system that was previously required. A civil matter, but corporations have warped law enforcement to work on their behalf also.

Then you get into licensing and permissions granted when you upload to a site. If you grant Imgur a non-revocable license to reproduce your imagery, for example, other social media posters that keep reposting on that site really aren’t infringing.

OpenAI does not propose to take unlimited licensing of GPT content for themselves. This definition statement (using capitalization) is the clearest:

GPT Content. The information that you include with your GPT (for example your GPT name and description) is your Content.

Service terms of OpenAI are mainly about what users or developers can and cannot do. Perhaps some of the higher-profile clauses that could initiate OpenAI action against an account (such a one noted to do mass copying):

Disallowed usage of our models

  • Coordinated inauthentic behavior
  • Plagiarism
22

I imagine today they are scrambling to have an AI check for duplicate creation of custom GPTs, and automatically flagging them

24

I’m not sure where you’re getting your information, but copyright protection is instantly active as soon as a creative work enters a tangible medium, i.e. a page, file, or a website.

This is regardless of whether you have a registered trademark or not.

Per Harvard.edu:

Under the U.S. Copyright Act, as soon as a creator sets a work ’down in a tangible medium (like drawing on paper, recording on cassette, video on tape, or words on a website), the creator has the exclusive right to:

  • To reproduce and make copies of the work;
  • To prepare derivative works based on the work;
  • To distribute copies of the work to the public;
  • For literary, musical, dramatic, and choreographic works, pictorial, graphic, or sculptural works (including individual images or screenshots from audiovisual works), to display the work;
  • for literary, musical, dramatic, and choreographic works and audiovisual works, to publicly perform the work; and
  • for sound recordings, to perform them publicly through digital audio transmission.

…and grants the copyright owner the right to sue for damages when infringement is committed.

And regarding the use of AI to generate works on your behalf,

but it’s not “open to question” if copyright applies when generative AI is used to create the work—Ownership (correction, it is open to question, as generative AI is not copyright protected without adequate human intervention)

…but rights to distribute, sell, etc are defined by the tool’s organization in their ToS, in OpenAI DALLE’s case:

Subject to the Content Policy and Terms, you own the images you create with DALL·E, including the right to reprint, sell, and merchandise – regardless of whether an image was generated through a free or paid credit.

Hope this clears up any confusion!

[Edit: Tried removing my comment when I noticed the thread’s age, apparently deleted posts aren’t actually “deleted” when edits can be viewed. So just keeping the post here, apologies.]

25

I also found out today that this website Nuclear Safety Virtual Assistant-Free Nuclear Safety Insights copied my custom GPT Nuclear Safety Virtual Assistant ChatGPT - Nuclear Safety Virtual Assistant . What can I do? Where I should report this?

I think in fact they are copying many of the OpenAI products and offering them for prices lower than the subscription for ChatGPT Plus Plan tarifaire YesChat.ai . What does OpenAI think about this?

26

You are 100% incorrect here.

I am getting my information from the United States Copyright Office.

https://www.federalregister.gov/documents/2023/03/16/2023-05321/copyright-registration-guidance-works-containing-material-generated-by-artificial-intelligence

You are 100% incorrect here.

An organization cannot grant you copyright to something which is not copyrightable.

You need to work on your close-reading skills my friend… You can own the image and have a right to reprint, sell, and merchandise an image without owning the copyright to that image.

For instance, I have the right to reprint, sell, and merchandise an image which is in the public domain.

As it stands now, under US copyright law, a work requires human authorship to be eligible for copyright protection. AI-generated works lack human authorship, ergo they are not eligible for copyright protection.

This has been well established for a very long time now. The most recent high-profile case being Kris Kashtanova’s “Zarya” from over a year ago.

Hope this clears up any confusion!

27

Unfortunately, there’s nothing you can really do as you granted them a license to use your content when you published the GPT.

28

This is a settled debate, and I’m honestly happy that is the outcome, or someone could become the biggest copyright troll by just write a script that generates images and publishes them to a website.

I’m kinda curious how much “human” you have to add before it’s copyrightable though, like my phone has various “AI filters”, I wonder if those images are copyrightable :thinking:

29

All of what’s been said is under the assumption that the appropriate level of human intervention had taken place to make the work eligible for protection under copyright law.

I’m not talking about people who are just pulling images directly out of AI and attempting to assume copyright protection. I’m specifically referring to professionals who leverage AI to assist in creating works and modify them further using professional graphics editing tools.

A human intervening with generated content and adequately modifying it to a final version would deem it eligible for copyright protection.

This is a commonplace for graphics professionals, particularly with AI-generation tools baked directly into design software like Canva and soon to be Adobe Firefly with Photoshop.

You just duplicated your message…

To clarify: An organization can’t grant copyright ownership to an item that’s been generated without any human intervention or authorship. But an organization who owns the tool can still privately impose restrictions on what a user can legally do with any generated images via their terms of service agreement.

They can still arguably prevent a user from assuming rights to reproduction and commercial use through the ToS agreement, just as they can grant those permissions to the user, even though there’s no copyright owned.

Not sure I follow what point this addresses in the discussion. I never implied that one couldn’t reproduce or redistribute works without owning their copyright.

I’m specifically referring to rights and protections granted by and within copyright law to works assisted with generative AI, so long as the requirements of adequate human authorship have been met.

And this time, I really do hope this clears up any confusion :zap:

30

In all honestly, I’m personally in the camp that supports copyright of AI generated content.

AI is just another tool and medium of creation, similar to a pen, typewriter, chisel, etc. AI doesn’t generate content entirely on its own, a human still has to initiate the process and provide input/prompting to mold the output into what they desire.

What’s silly is that I could write a script in JavaScript to generate an image on an HTML canvas using a few mathematical and random functions to create a unique image each time, and the output would be arguably considered copyright protected.

Why? I used a machine and automation to create the image.

A human still has to interact with an AI image generator to specify its output, that’s authorship.

31

It’s really hard for me to read this,

and come away understanding you meant this,

Perhaps you should try to be more clear with the ideas you are trying to communicate.

Then you probably should have articulated that in the first place.

That “adequately modifying” bit is rather important. It also has not yet been quantified and there is no clear line when it’s been met. To the best of my knowledge there still hasn’t been a clear case of a work initially created by a model then later modified by a human which has been properly registered in the United States.

Then you probably shouldn’t have written this,

Where you directly conflated copyright with OpenAI’s terms of service for DALL-E—a service which exclusively generates images without human authorship.

Because you conflated those rights with owning the copyright.

Look, there was no confusion. Everyone understood exactly what you wrote as you originally wrote it.

If you intended us to understand something other than the words you wrote, you should have written different words.

34

Look, you are the one who revived a four-month-old topic by posting factually incorrect information then, when corrected, instead of simply accepting that you tried to reframe what you wrote as being correct if we made a bunch of assumptions not in your original post.

I’m closing this topic as it is stale and has veered off-topic since being revived.