Generative AI Commits Intellectual Property Crime – It’s Time To Call It Out. And Fix It With Legislation.

A few days ago, I was listening on radio to an artist who was talking about how her experience of making her art was being destroyed by the generative AI rip-offs of her work, creating fake digital versions of her paintings which were then available to anyone for free. The interviewer was so sympathetic.

When creative people (of all kinds) fall into the “I’m creative and it hurts” trap, it sets back the argument, which is a very strong one, for the validity of copyrights and the outrageous and complete violation of copyright that generative AI employs today.

The fundamental problem for AI, as for the entire digital era, is copyright: the right of the creator to their original works and complete legal control over who can copy and under what circumstances. AI is designed to take unfair advantage of the original and creative works of others (scientific, visual, conceptual, mathematical, and so forth) without either attribution or payment.

If anyone wondered why the writers and actors are on strike, this is why.

Creative works (in whatever the medium they appear) belong to their authors. This is an irrevocable human right, recognized the world over.

What does a copyright mean?

When a creator of any kind creates a thing, the creator has the full and complete monopoly rights to what they created. A common-law copyright exists even when the symbol ? is not present, but of course adding this mark increases the ability of the rights holder to enforce their right to not have their intellectual property duplicated (and is a prerequisite to registration, which is how you can sue infringers). This law has been in place in the United States since 1790, with many updates and clarifications in the centuries since to address the evolving nature of authorship and creation.

The problem for creators is that copyright is enforced in civil court – that is, the creator must sue the copyist with all the attendant costs and challenges that entails. Copyright cases can achieve significant notoriety however often a legal letter to an infringer is enough to cause the infringement to be stopped, and some form of damages to be paid.

The Legislation Creators Need Now

The fastest way to secure the control of our global intellectual property in the face of AI “competition” is for our legislators to create laws that require that anything created by an AI engine identify each and every source for the content: The web page. The author. The story title. The paragraph. This could be footnoted into the content created by the generative AI engine, with the footnotes tying to the content areas.

For example, if the engine identified that “House cats like to sleep most of the day” it could have as many footnotes as the sources used to come up with the statement:

House cats like to sleep most of the day.(Sources in footnotes: 1,2,3,4,5,6,7)

It must be required that every source used to create the AI text be included with the text, so there is no uncertainty about where the ideas came from, and to credit the creator for their work.

And then, through an agency that would function somewhat like the American Society of Composers, Authors and Publishers (ASCAP, https://www.ascap.com/), a small royalty can be paid back to the contributors to the idea. The financial technology exists to perform microtransactions such as these (it’s happening now with Apple Music for example), and the revenues for the use could be stockpiled in the event the author/creator was not listed with the agency – though once such a registry was available, there would be a massive rush to register in order to claim all financial proceeds from the use of a creator’s work.

This could involve the development of a modern equivalent of the Digimark or other identifying symbol. Today we have ? to indicate that content is copyrighted and that is a good place to start; images may need some more complex digital data to be embedded in them which the AI engine is required to look at and add to the final content.

Isn’t AI content mining “Fair Use”?

The use of intellectual property by a generative AI engine is not a fair use. The doctrine of fair use permits brief excerpts to be used under very proscribed circumstances. Section 107 of the Copyright Act is explicit as to what is fair use: “criticism, comment, news reporting, teaching (including multiple copies for classroom use), scholarship, [and] research,” which allow use without permission or payment of a royalty. This is not a broad approval for an AI engine to mine, steal, and take credit for the creative and intellectual work of others for profit – and it has been bandied about a reason that AI can operate with impunity. This is well-tested law. AI is violating this law.

The Digital Millennium Copyright Act[1]

From Wikipedia: “The Digital Millennium Copyright Act (DMCA) is a 1998 United States copyright law that implements two 1996 treaties of the World Intellectual Property Organization (WIPO). It criminalizes production and dissemination of technology, devices, or services intended to circumvent measures that control access to copyrighted works (commonly known as digital rights management or DRM). It also criminalizes the act of circumventing an access control, whether or not there is actual infringement of copyright itself. In addition, the DMCA heightens the penalties for copyright infringement on the Internet. Passed on October 12, 1998, by a unanimous vote in the United States Senate and signed into law by President Bill Clinton on October 28, 1998, the DMCA amended Title 17 of the United States Code to extend the reach of copyright, while limiting the liability of the providers of online services for copyright infringement by their users.”

The last sentence is key. The DMCA offered a statement of protection, but did not hold the purveyors of services that allow copyright infringement to take place to be held accountable. At the time this was passed, the creative community was very unhappy – because it became nearly impossible to hold anyone accountable when the digital environment was basically about copying. However, as we look at the AI application to the DMCA, this is less clear. There is absolutely no reason why a generative AI engine could not have software inside it to identify the sources, and when copyright is exerted by the intellectual property owner, to contribute to a system to provide revenue back.

What about search engines?

This is a classic whataboutism. If the AI search engine must return all the resources, the seeker will be much more able to identify sources and validity of the data being provided. This especially can help for scientific questions where conspiracy theorists have muddied the online waters with false information – which an AI engine will not discern – and indeed has been shown to amplify in disturbing ways. And of course, if the data is used, the rights holders can claim their fair share of revenue from the use of their intellectual property – whatever type it is.

Oh this is all too hard – why bother?

As we all know, the AI engines today are stealing the work of others, blending it together, and spitting out a version of the world that very often increases confusion or perpetuates false ideas, as well as ripping off the rights holders. At its worst, it spits out homogenized ideas that bring little insight – and ignore the flashes of insight by exceptional thinkers. These engines do not create. They require humans to create, which they then use and abuse. As human creators receive less and less income and recognition for their works, that creative resource has risk of fading away – until we live in a homogenized world of AI pablum and risk forgetting what maverick creativity can do for our society, our technology and our planet. That’s not the future I want to live in. And neither should you.

#FightForCopyright #AIStealsFromCreators #EnforceCopyrightsInDigital

? 2023 Cynthia H. Batty. May be shared. All rights reserved.

[1] https://en.wikipedia.org/wiki/Digital_Millennium_Copyright_Act