and here comes South Indian Elvis.

(Vivek laughing)

- [Narrator] This is Reface, an app that lets users

swap faces with others using AI,

and it's currently embroiled in a lawsuit.

It's one of dozens of AI related lawsuits

filed in the last couple of years.

- I'm gonna break down three

of the biggest cases at the intersection of AI

and intellectual property law.

Maybe the biggest issue facing copyright lawyers today is

whether or not works created using generative AI tools

are even protectable?

The case that illustrates this copyright conundrum

perhaps the best is Thaler v Perlmutter.

- [Narrator] In 2018, scientist Dr. Stephen Thaler

applied for copyright protection

for this piece of art, which he said was produced

by a generative AI system he created.

The Copyright Office rejected his application

saying creative works must have human authors

to be copyrightable.

- And so what the Copyright Office is basically saying is

that we're not gonna extend copyright protection to works

that are made by machines.

- [Narrator] Thaler sued over the rejection and lost,

but he didn't give up.

He appealed the ruling and a decision is expected

by the end of this year.

The Copyright Office declined to comment.

- In order to receive copyright protection, a work needs

to be original, fixed to a tangible medium,

and have human authorship.

This element of human authorship is not really disputed,

discussed or litigated very often.

However, there was a case that got some headlines

that did focus on this element of human authorship.

- [Narrator] In 2011, wildlife photographer David Slater

set up and left his camera unattended

and a monkey named Naruto allegedly took photos

of himself with Slater's camera.

- PETA argued that Naruto, not Slater,

should be the lawful owner of the copyright.

Ultimately, the federal court said

that photograph lacks human authorship

because Naruto is a monkey.

- [Narrator] And the rise of AI raises new questions.

If a human edits artwork made by AI,

can that count as human authorship?

- Maybe The first case

that actually spotted this copyright conundrum involves

the artist and computer scientists named Kris Kashtanova.

- [Narrator] Kashtanova used the generative AI

tool Midjourney to create the artwork

in their graphic novel.

They said they personally edited some

of the artwork using Photoshop,

but the Copyright Office said the edits were too minor

to be entitled to copyright.

The Copyright Office also said in its letter to Kashtanova

that if there were a sufficient amount

of original authorship made by a human

that could warrant copyright protection.

Kashtanova's attorneys said in a statement

that the Copyright Office's decision was incorrect

and is unsustainable.

- So why is all of this such a big deal?

Without the ability to get a copyright registration,

people who are making work in this way

will not be able to meaningfully commercialize that work.

I think either a federal appellate court is going

to expand on the definition of

what human authorship means,

or I think Congress will be lobbied to a point

where they will need to define again, the meaning

of human authorship as it may be relates

to contemporary society.

Why is the intersection of AI

and right of publicity laws important right now?

AI as a technology has made it exceedingly easy

to co-opt, or repurpose somebody's name, image,

and likeness for some kind of commercial purpose.

♪ I came in with my ex like Selena to flex ♪

Yeah, I mean, that sounds like Drake.

- [Narrator] But it isn't.

It was made by an anonymous producer named Ghostwriter

who used generative AI tools to fake Drake

and the Weekend's Voices.

Ghostwriter's song hard on my sleeve had 600,000 plays on

Spotify before it was pulled down following a copyright

infringement complaint from Universal Music Group,

and it isn't the only example.

Kyland Young a former contestant on CBS's Big Brother.

- This is a fun opportunity.

We come back to more friends.

- [Narrator] Sued Neocortex, which owns the app Reface

we saw earlier.

He said his face was in the app's library for paying users

to edit without his consent,

which violates his right of publicity.

Young's complaint says that the use of his image

for commercial benefit is the core of the issue.

- One example of this is a 1992 case involving Samsung.

When they ran this ad.

I probably don't even have to tell you who that is,

and that is the problem.

- [Narrator] The robot in the print ad is meant

to look like Vanna White, a host on the game show

"Wheel of Fortune."

In 1992, she sued Samsung for violating

her right of publicity,

and the court agreed.

That's the question that Kyland Young's lawsuit

is wrestling with.

- I think Neocortex is going to have a very difficult time

defeating this claim by Young.

- [Narrator] Reface said Young's images are no longer

available in the app.

The company argues that

because the app's output is derivative,

the Copyright Act preempts Young's right of publicity claim.

- Most of the actual lawsuits relates

to this thing called training data.

- [Narrator] Or the material used to create large language

models that power generative AI systems like ChatGPT.

One of the first cases involving training data was filed

by a group of artists in 2023.

- What these artists are saying is

that their copyrighted works,

which live in digital form online, are used

and scraped to train the large language models used

by Midjourney, Stability AI, and the other defendants.

Since these artists never licensed their work

to these defendants, they argue that that use

of their copyright work violates their copyright rights.

- [Narrator] They also argue that the artwork produced

by these AI systems should be considered copyright

infringement because of how similar

they look to the originals.

- Substantial similarity is the legal standard used

by courts to determine whether

or not copyright infringement has occurred.

Courts will usually focus on the similarities,

not the differences,

and they'll focus on the similarities

of the protectable elements among each work.

- [Narrator] Like the subjects in this Getty image

and the angle of the camera, Stability AI didn't respond

to requests for comment.

- I think that at the end of the day,

the training standing alone is probably going

to be okay, and here's why.

There is a case from a few years ago involving HIQ

and LinkedIn, HIQ was developing some recruiting software

by scraping publicly available data from linkedin.com,

and what the court ultimately said in

that case was scraping publicly available

information is okay.

These platforms are still going to have to ensure

that the output does not constitute copyright infringement.

- [Narrator] There will likely be more of these copyright

issues, especially when OpenAI releases its text to video

tool Sora to the public.

- You can imagine now if somebody wants

to create a cityscape of like Tokyo or New York City

or Paris, right, they might end up

actually creating modified

but recognizable versions of famous trademarks.

- [Narrator] OpenAI declined to comment.

The law is just beginning to catch up

to the fast advancement of AI.

- Technology in general moves very fast

and tests the bounds of the existing laws that we have.

Oftentimes, it's these types of technologies

and shifts in technology and culture

that force us to rethink the law

or reapply the law, or ultimately even rewrite the law.

(gentle music)