A Fair Content Licensing Solution for Generative AI

A Realistic 3-Tier System That Works for Both Media & Tech

Good morning everyone — it’s time for your weekly brAIn dump! This week’s “mAIn event” features my proposed fair licensing solution for generative AI training on copyrighted works. Then, it’s the “generative AI video of the week” (trust me, you’ll want to see this one …). Finally, it’s the “AI Litigation Tracker” — updates on key generative AI infringement cases by McKool Smith.

But first …

Last week, Shira Perlmutter, director of the U.S. Copyright Office, told a U.S. Senate oversight panel that the issue of “fair use” in the context of generative AI is “the most controversial of all the areas that we’re addressing in the [AI] report” that her office will publish at some point in 2025. This is how she characterized what’s at stake: “the copyright owners are saying it cannot be fair use to take everything that we’ve produced through a lifetime of training and inspiration and to use it as raw material to create content that will compete with us in the marketplace.”

---

I. The mAIn event: a fAIr Licensing Solution for Generative AI Training on Copyrighted Works

A couple weeks back, I proposed a fair, workable licensing solution for generative AI companies to be able to train their LLMs on copyrighted works — both ethically and lawfully. But my solution was buried in my larger discussion of the content licensing “budgets” of each major GenAI developer (which you can access here via this link).

Given the level of interest (and there was understandably a lot of it), I wanted to give my solution the spotlight it deserves with the goal of constructively moving the industry forward. After all, it’s in everyone’s interests to take away the “friction” of acrimony and uncertainty when it comes to the current state of play where there has been massive unlicensed training — which inevitably leads to equally massive risk and wasteful litigation. Why not instead move forward together to share in the value created by the great generative AI “unlock?” It’s time to do “fair” licensing deals now. So, let’s go!

My Proposed 3-Tier “Opt In” Solution

I propose a 3-tiered “opt in” licensing solution for rights-holders that compensates them for past unlicensed training — and pays them for inclusion of their works in next-generation LLMs developed during a defined period of time (i.e., the term of the licensing deal):

(1) Tier 1:

1-1 Direct Licenses Between Individual GenAI Developers & Individual Major Rights-Holders (media and entertainment companies with significant content libraries). OpenAI’s $250 million deal with News Corp is an example;

(2) Tier 2:

1-1 Direct Licenses Between Individual GenAI Developers & Aggregated Rights-Holders (individual media and entertainment companies with smaller, yet still valuable, content assets — who “pool” their assets together to bring content scale and diversity). This is a “strength in numbers” approach; and

(3) Tier 3:

Automated “Opt In” Platforms for All Other Individual Rights-Holders (smaller rights-holders who want to participate in GenAI monetization — in a way that is akin to music public performance licenses). Several VC-backed companies are on a path to enable this kind of “automated” approach. I know several of them.

So What’s The Right Number?

So, the next question becomes how much GenAI developers should pay rights-holders in these licensing deals for use of the essential content they need (without which much of their GenAI tech investments would have little commercial value)?

To answer that fundamental question, maybe we should at least take a look at how Big Tech and Big Media solved this analogous question in earlier massive Tech-tonic shifts. Let’s take the last major tech-fueled disruption in the world of media and entertainment — streaming. YouTube pays out about 55% of its ad dollars to rights-holders. Meanwhile, Spotify pays out about 70% of its revenues to music rights-holders.

Maybe those kind of streaming percentages don’t work here in this GenAI context due to the massive capital investments necessary for generative AI to “do its thing.” At least arguably, streaming also monetizes the relevant licensed content more directly. Each content source used for GenAI training is typically less directly “traceable” to the generated outputs (although, to be clear, that reality doesn’t change the fact that the unlicensed use is infringing) (I’ve written about that time and time again, including in this article here where I lay out three possible paths to infringement).

In any event, when we objectively look at the relevant estimated total GenAI-related content licensing expenditures to date — which range from a mere .33% to 5% of total GenAI expenditures by the major GenAI developers (per my detailed analysis here that I strongly urge you to check out) — it’s hard to justify those as being the “right” or “fair” answers given the critical role content plays in GenAI value creation.

So, what is fair? That’s all playing out in real time right now.

The Trump Effect?

I’ve been asked whether Trump’s recent election changes these GenAI content licensing dynamics. While I certainly expect AI “Trump Change” — policy shifts that slow the pace of AI-related legislation and regulation (much to the delight of Silicon Valley’s AI “Acceleration-ists”) — none of this should slow the pace or size of GenAI-focused content licensing deals. And it certainly won’t slow down the pace of litigation amongst the major generative AI developers and media rights-holders.

I want to hear from you. Send me your own ideas and feedback to peter@creativemedia.biz.

---

II. And speaking of the election, here’s your GenAI video of the week … just watch via this button …

… too soon???

---

III. AI Litigation Tracker: Updates on Key AI Infringement Cases (by McKool Smith)

Partner Avery Williams and the team at McKool Smith (named “Plaintiff IP Firm of the Year” by The National Law Journal) lay out the facts of — and latest critical developments in — the key generative AI/media infringement cases listed below via this link to the “AI Litigation Tracker”.

(1) Raw Story Media v. OpenAI (about which I wrote at length last week via this link)

(2) Dow Jones, et al. v. Perplexity AI (about which I wrote at length a couple weeks back via this link)

(3) The New York Times v. Microsoft & OpenAI

(4) Sarah Silverman v. OpenAI (class action)

(5) Sarah Silverman, et al. v. Meta (class action)

(6) UMG Recordings v. Suno

(7) UMG Recordings v. Uncharted Labs (d/b/a Udio)

(8) Getty Images v. Stability AI and Midjourney

(9) Universal Music Group, et al. v. Anthropic

(10) Sarah Anderson v. Stability AI

(11) Authors Guild et al. v. OpenAI

(12) The Center for Investigative Reporting v. OpenAI

NOTE: Go to the “AI Litigation Tracker” tab at the top of “the brAIn” website for the full discussions and analyses of these and other key generative AI/media litigations. And reach out to me, Peter Csathy (peter@creativemedia.biz), if you would like to be connected to McKool Smith) to discuss these and other legal and litigation issues. I’ll make the introduction.

---

And check out my firm Creative Media and our AI-focused services

---

Send your feedback to me and my newsletter via peter@creativemedia.biz.