Generative A.I. Authorship or Authorship through Generative A.I.: The Copyright Dilemma | PhotoVogue Festival 2023: What Makes Us Human? Image in the Age of A.I.

I m honored to be called to contribute

to the debate on generative AI.

My task today is talking about authorship of AI

or through AI, and therefore, the intersection

between copyright and generative AI.

This topic became a hot topic immediately after the release

of ChatGPT-3 in November 2022,

so exactly one year ago.

And nowadays, the question is whether a work

entirely generated by a generative AI model

can be considered an artwork

and therefore enjoy the copyright protection,

and in the affirmative case, who is the author?

[paper rustling]

So...

It maybe hard to believe, but there was a time in the past

when photography was viewed with prejudice and suspicion

by the art world, like digital art may be seen today.

Many believe that because photography involved chemicals

and machinery instead of human and spirit,

it could not be considered on par with drawings

or a painting.

Instead, photography was seen closer to fabrics

being mass produced by machine.

In 1859, Charles Baudelair said, wrote, sorry,

Photography is the refuge for those who aspire

to be painters but lack the talent or are too lazy

to complete their studies.

The reliance on photography will further diminish

the already limited French artistic genius.

[paper rustling]

But [chuckles] the question of the classification

of photography as a copyrighted work

and expressions of the author s creativity

was first discussed in the United States

in the Sarony case in 1882.

For the first time, the Supreme Court of the United States

was called to answer to the questions

of whether a pictures can be qualified as an artwork.

And the Supreme Court came to the positive conclusion,

Because the picture comes from the photographer s

own original mental conception

to which he gives visible form, said the court.

By posing the subject in front of the camera,

selecting and arranging the costume,

arranging and disposing the light and shade and so on.

These findings show, said the court,

that the picture is an original artwork,

is the product of the photographer intellectual invention.

The Supreme Court added that,

The author is the person who effectively

is the cause of the pictures which is produced.

The authorship involves originating, making, producing,

and inventing the thing which is to be protected,

and therefore, photography as well as drawing and painting.

The ruling is still...

This ruling is still quoted today as the pivotal ruling

on the protection of photography as an artwork,

and it is also been cited as a fundamental precedent

in the examination of the question of whether a work

generated entirely by an artificial system

absent human involvement should be eligible for copyright,

does it falls under protection of copyright law

upon its creation?

[paper rustling]

We are approaching new frontiers in copyright

as artists put AI in the toolbox

to be used in the generation of new visual

and other artistic works.

In particular, generative artificial intelligence

describes algorithms that can be used to create new content,

including audio, code, images, text, et cetera,

and recent breakthroughs in the field have the potential

to drastically change the way contemporary artists

approach content creation.

The increased attenuation of human creativity

from the actual generation of the final work

will raise challenging questions regarding

how much human input is necessary

to qualify the use of an AI system as a...

To qualify the user, sorry, of an AI system

as the author of the generated work,

the scope of the protection obtained

over the result and image,

how to assess the originality of the AI-generated works,

and how copyright might be used

to incentivize creative works involving AI.

[paper rustling]

Very recently, the United States District of the Washington

faced with part of these questions and answered that,

Work entirely generated autonomously by a computer system

is not eligible for copyright protection.

This is the judgment in Stephen Thaler

versus US Copyright Office issued in August of this year.

And the artwork not allowed to enjoy copyright protection

was titled A Recent Entrance to Paradise.

Under US Copyright Act, the copyright protection attaches

immediately upon creation of original works of authorship

fixed in any tangible medium of expressions,

provided that those works meet certain requirements.

In particular, copyright work is limited

to original intellectual conception of the author

as a human being.

Because of these two consequences,

The author of a piece of art must be a human being,

not an algorithm or a machine.

And second, The absence of any human involvement

in the creation of the work,

the work entirely generated by a computer system

is not eligible for protection by copyright.

In particular, the Supreme Court,

using the Sarony argument that we saw before, reminded that,

The recognition of the copyrightability of a photograph

rested on the fact that the human creator, not the camera,

conceived and designed the image

and then used the camera to capture that image.

Because of that, the photographer is the author

of the product of his own intellectual invention.

Based on this argument,

but without going through an in-depth analysis

of the technology involved, the Supreme Court affirmed

that an AI autonomously generated work

with no human involvement

is not eligible for copyright protection.

[paper rustling]

This decision is aligned...

Next.

Is aligned with what s stated by the US Copyright Office

in the recent Copyright Registration Guidance

related to works containing material generated by AI.

In this document, Copyright Office,

going a little bit more in-depth into the analysis

of the technology involved, affirmed that,

The work lacks human authorship

when an AI receives solely a prompt from a human

and produces complex written, visual, or musical works

in response.

In this case, the traditional elements of authorship

are determined and executed by the technology

and not the human user.

Based on the Office s understanding

of the generative AI technologies currently available,

users do not exercise the ultimate creative control

over how such system interpret prompts

and generate material.

Instead, these prompts are aimed to identify

what the prompter wishes to have depicted,

but the machine determines

how the instructions are implemented in its output.

When an AI technology determines the expressive element

of its output, the generated material is not the product

of the human authorship.

As a result, the material is not protected by copyright

and must be disclaimed in a registration application.

Based on this reasoning,

the US Copyright Office refused protection for images

in a graphic novel generated

using the AI platform Midjourney

and titled Zarya of the Dawn.

In other cases, however, says the US Copyright Office,

A work containing AI-generated material

might also contain sufficient human authorship

to support a copyright claim.

For example, a human may select

or arrange AI-generated material

in a sufficiently creative way

that the resulting work as a whole

constitutes an original work of authorship,

or an artist may modify material,

originally AI-generated material, to such a degree

that the modifications meet the standard

for copyright protection.

To sum up, the US copyright policy does not mean

that technological tools cannot be part

of the creative process.

Authors have long used such tools to create their works

or to recast, transform, and adapt

their expressive authorship.

In each cases, what matters is the extent

to which the human had creative control

over the work s expressions

and actually formed the traditional elements of authorship.

[paper rustling]

If we jump on this side of the ocean,

we can appreciate the approach of the European countries,

and especially Italy, to the topic of the copyrightability

of work generated by AI is not very dissimilar.

Indeed, according to principle

of international copyright law and EU law,

a work is protectable under copyright

when it is the fruit of the mind

and possesses the creative character.

It is expression of the author s own intellectual creation.

This does not mean that contents generated by AI

directly follow into public domain

only because they are caused by an algorithm.

The key point is the human involvement or contribution

into the creative process that contemplate generative AI

as a tool used to such a creative process.

In a recent decision, the Italian Supreme Court

has been called to determine if a software-generated work

might be protected by Italian copyright.

It wasn t a work generated by AI,

but the reasoning can be easily extended

to an AI-generated work.

The court with few lines summed up

what we have been discussing up until now.

The Supreme stated that it is mandatory

to go through a factual assessment

to verify whether and to what extent

the use of the software absorbed the creative elaboration

of the artist who had use it.

In other words, the possibility that the output

of the generative AI model

is equipped with the requirements of creativity

and therefore is protected by copyright.

It is not excluded a priori,

but it will be necessary to verify that the author

has not completely delegated the responsibility

for the creative process to the machine or algorithm.

[paper rustling]

But the issue of the measuring

of human creative contribution

in the generative process of digital work

will be crucial, therefore, whenever it comes

to trying to invoke authorial protection

for works generated by or with the help of AI models.

It is likely that the results

of the generative artificial intelligence models used today

cannot easily aspire to copyright protection

if they are the result

of the [indistinct] decision making process

of the algorithm underlying the AI model,

with minimal input from the user,

which result in basic textual indications.

On the contrary, copyright protection could be invoked

if the user or the AI platform is able to prove

that the artificial intelligence model represented a model

or a tool within a more complex

and sophisticated creative process,

but which is the measure

or it is the degree of creative contribution

that allows the user of a generative AI model

to claim the copyright protection

over an AI-generated output.

Again, the US Copyright Office has tackled recently

this topic for the first time,

again with reference to an image created by Midjourney

that had won the first prize

in a Colorado national competition.

In fact, the author question, Mr. Jason Allen, sorry,

had applied for copyright registration

of an AI-generated work naming his himself as the author,

the one depicted there.

The US Copyright Office refused once again

to register this work because it failed

to meet the de minimis standard for copyright protection

and the author has failed to disclose

and disclaim the AI-generated portion of the work.

The applicant argued that he contributed significantly

to the creation of this image.

His creative inputs included providing

around 642 text prompts and revisions of text prompts

and using software to remove flaws and upscale the image.

Nevertheless, the US Copyright Office said

this was not enough.

The step in the process were ultimately dependent

on how the Midjourney system processed Mr. Allen s prompts.

No matter how creative the inputs were,

the output was the result of interpretation by Midjourney

based on its training dataset.

The US Copyright ruled that prompts

are not direct instructions,

which Midjourney understand like humans do.

For the moment, therefore, content generated solely

through a generative AI model

cannot aspire to authorial protection

and necessarily fails into the public domain even when,

based on the strict US Copyright Office approach,

it is conceived by the mind of the user or the AI platform

who gives shape to the idea he has conceived

through a technological process.

Does this mean that third parties can freely reuse, modify,

and rework the content?

The answer is yes, at least so far,

unless the measurement of the creative contribution

of the user of the AI platform

allows us to conclude otherwise.

But the unit of measurement

based on which this creative contribution is measured

is entrusted to the discretion of the interpreter.

[paper rustling]

But an alternative solution is perhaps conceivable,

a solution that does not require

to go through the discretional measuring

of human contribution each time a work generated by AI

aspires to enjoy copyright protection.

I m referring to the drafted bill,

proposed in French in September of this year,

aimed at regulating the relationship

between copyright and artificial intelligence.

The approach is inspired by the precautionary principle,

given that generative AI is represented

as free to fragile sectors of economy,

those devoted to artistic production.

Among other things, the bill provides

that when an AI-generated work is generated

based on preexisting identifiable contents,

the author of the AI-generated artwork are the authors

or holders of the exclusive rights

although the work that made the creation

of the AI-generated content possible.

Companies representing authors

or collective management organizations

would collect and redistribute the related compensation.

On the other hand, when an AI-generated work is generated

based on works that are not identifiable,

the most likely scenario,

the author might be the user,

and a tax would be established for the benefit

of the copyright collecting management organizations

in charge to redistribute to all authors

that contributed with their works

to the training of the AI model,

and such tax would be imposed on the company

that is operating the AI system

used to create the said artificial work.

I don t know if the French approach is feasible,

technically feasible under a low perspective,

and even if the most correct.

I don t know if the most correct.

Certainly there are pro and cons

depending on the interest of different scenario,

but the interesting thing is that in this case,

the bill says that it is not possible to think of a system

in which the output of generative AI models

is protected by corporate legislation

and therefore the content is economically exploitable

without going through the recognition of compensation

to the authors of the works

that have been used to train the same AI model

that produced the content on which some form of exploitation

is expected to be exercised.

Thank you very much.

Hope it was pretty clear,

and I leave the floor to the next speaker.

[audience clapping]