WIPIP Panel 3: Deepfakes, Celebrities, and Movies

A Digital Right of Publicity for the AI World (Emma Perot)

Prehistory: ROP covers lookalikes, soundalikes, video game avatars
(at least for realism).

Persona as training data. Theories of personality: users
informed about use; many social media companies do not allow opt-in or opt-out;
performers can be subject to exploitative terms. Unjust enrichment: in absence
of compensation for training and licensing to third parties. Incentives
undermined, though this won’t apply to the average social media user. We tend
to favor art we think is made by humans. W/o proper protection we’ll undermine incentives.

Input stage: is consent obtained? Maybe, sort of, via terms
of service, contracts with publishers, agreements made pre-AI.

Output stage: could be identifiable or unidentifiable. The
latter isn’t really a problem for ROP. But the ability to create unidentifiable
persona makes input approach important—persona has been used for the training.
Risk of nefarious uses already addressed in many situations; but expressive
uses could be covered which could interfere with realistic depictions, e.g. in
biopics. If you’re conveying information about the person perhaps we want to
allow that (as opposed to casting them in a role).

No FAKES: positives—exception for satire/parody, disclaimers
not permitted. Negatives: doesn’t address training, uses to create composite
figures, licenses last too long (10 years). SAG-AFTRA agreement: more positive
b/c TV/theatrical uses they say: each capture and use of digital replica should
be consented to and compensated.

Need clarity on how consent for input is sought and how
outputs will be treated in expressive works. Avoid labor displacement.

Betsy Rosenblatt: Different interests from people who make a
living from their persona v. someone who is more concerned w/privacy.

Lisa Ramsey: theories of persona is one framing, but what
are the goals of the law is another. What are you trying to protect? Right to
make $ from performing? Right to stay out of database?

[Persona is a bad fit for a consent-to-train framework because
the details of a person’s life etc. are facts. You may be thinking about songs
and movies, but the NYT wants to be able to license its news stories w/o being
overridden by the subjectss. Moreover, persona is constructed with others. When
I tell my life story, my husband’s life story is inherently implicated.

Incentives: if you are relying on the preference for human-made
work, what you want is a disclosure regime; licensing by the person will lead
to lots of deception.]

Transformative Celebrity (Rebecca Curtin)  

ROP claims might be implicated by methods users use—Zarya of
the Dawn looks a lot like Zendaya, and that’s b/c the prompter used her name to
produce a character who resembles Zendaya. Use of names in prompts alone, w/o
reference to whether an infringing result is produced, has been raised by artists
who alleged that allowing users to request art in the style of X was a
violation of the ROP.

Transformative use was imported into ROP from ©, but © has
recently devalued new meaning and message. Should ROP follow ©? Should © return
to focus on new meaning, especially for use of persona? Does it make sense for ©
and ROP defenses to differ?

Transformativeness is not the worst approach of the
possibilities in ROP.

Warhol’s commerciality test sounds a lot like the
predominant use test in ROP cases (yuck).

Thinks Griner v. King was wrongly decided by 8th
Cir.—use of “success kid” meme was not fair use even used in political speech.
Inherently expressive way of using image as cultural reference, but court
reasons that controlling the commercial use of the template was the point of seeking
© and thus not transformative—devalues what the politician was trying to say. Warhol
is undervaluing new message & meaning, and expanding that approach to ROP
would be devastating to the transformative power of celebrity itself.

Charles Duan: the speech is out there; it’s just a question
of who gets paid/the rents.

A: chills speech by future potential speakers.

RT: experience w/state dilution law suggests courts are
neither willing to use nor capable of using two different tests with the same name.
Return to productive use, but this time for the ROP?

Trademark Law’s Trouble with Titles (Stacey Dogan)         

What would have happened if Barbie, the movie, was made
by a truly subversive entity instead of by Mattel? TM handles titles very badly
in ways that are becoming more problematic after JDI. Alone in the Dark lawsuit—based
on video game suing over title of movie Alone in the Dark, one of many films
under the same name.  Several courts in
the wake of JDI have concluded that titles may be source indicating, meaning
that Ds have to go through the regular LOC standard, which is much less
speech-protective. But these courts have gotten it wrong.

Titles and TMs do have some things in common. But
differences in purpose function and impact on audiences versus impact on consumers;
differences in consequences of extending legal rights to titles (speech impacts)—which
should mean different legal rules. Reference to titology, literary theory of
what titles mean.

Looks at different contexts: titles as TMs, recognized by early
20th-c courts as basis for unfair competition claims, but rightly
with a high standard requiring explicit misleadingness; Rogers and titles as
alleged infringement; Dastar and its conception of TM’s relationship to
expressive works (TMs identify source of physical object; titles are not TMs in
this view, but indicate a relationship b/t author & expression); and then JDI
and Kagan’s understanding of “use as an indication of source.” This should be
understood narrowly to preclude the treatment of titles as TMs.

Goal: easy resolution of most cases. Presume titles are not
source-indicating uses, which means that Rogers-like speech protection is
appropriate. But Rogers needs an update: can’t later claim TM rights based on
expressive uses. Some kind of estoppel, though implementation is tricky.

How to draw the line b/t mere use in title and other uses?
Advertising of title/work should clearly be allowed, but what about merchandise?
Advertising in connection w/merch?

Ramsey: would only apply rules against people who asserted
TM rights by applying for registration, claiming in a pleading, or maybe using
TM symbol—treat as omission. But what about Harry Potter and the ___? If there’s
no inherent meaning before adopted, maybe allow TM, but something like Alone in
the Dark already had meaning and thus shouldn’t ever be protectable.

A: two different forms of expressive use: one set of terms
involves common meaning like Alone in the Dark; also important to be able to engage
in nominative fair use-like titles like Barbie Girl.

Rosenblatt: how much do we need title exceptionalism to get
to this place? Dastar: relationship b/t title and content is not a source
relationship. We can understand that some titles are descriptive, generic,
arbitrary and approach them that way. (I don’t think arbitrary titles are
marks! Use as a mark is a separate requirement from placement on Abercrombie.)
But she’d limit that to rights claims; whereas when titles are the source of
alleged infringement we should treat that differently.

A: we have title exceptionalism on the protection side and
should retain it b/c of the expressive cost of granting rights. Titles are
creating more problems than characters—most of the Rogers character cases are
easily resolved in favor of the defendant b/c courts have found that use w/in a
work is more purely expressive.

FIRST: Archival Encounters That Set History in Motion
(Claudy Op den Kamp)

Videographic presentation of history of first film ©
registration in US.

from Blogger http://tushnet.blogspot.com/2026/02/wipip-panel-3-deepfakes-celebrities-and.html

This entry was posted in Uncategorized and tagged , , , . Bookmark the permalink.

Leave a comment