August 5, 2024 • 32 mins
In our next episode, we talk with top attorneys Ashima Dayal and Ina Scher from the New York City-based law firm of Davis+Gilbert. We’ll look at some of the legal landmines you need to avoid in marketing and advertising, including the intersection of AI and the Law.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:11):
(Transcribed by TurboScribe.ai. Go Unlimited to remove this message.) Welcome to Burning Questions from 500 Degrees.
We work together as a company, 500 Degrees
is a content fulfillment shop.
We have offices in Toronto, Miami, Columbus, Ohio,
and we work with some of the best
QSRs in the country, convenience stores and new
(00:31):
clients in the automotive industry, so we're expanding.
And we create thousands of images and assets
that are deployed on multiple digital platforms.
And if you wanna learn more about us,
go to weare500degrees.com and give us a
shout and we'll get back to you.
The best thing about my job is the
(00:52):
people that I come into contact during our
work day.
And it's great vendors, great associates, great people,
great law firms.
And that's what we wanna share in this
podcast.
And today we welcome the folks from Davis
and Gilbert.
I'm gonna call you guys the dynamic duo.
(01:14):
I've been working with Davis Gilbert for over
25 years and you guys have always been
great stewards of our legal issues and keeping
us out of trouble and just keeping us
on the path from contracts to trademarks to
whatever it is that we have to do
from claims.
I'm gonna first introduce Ashima Dayal.
(01:35):
And Ashima, you work on the counsel side,
I would call it, in terms of trademark
clearance, claims clearance, IP protection, numerous activities on
your side, Columbia Law and a background in
art history.
You've been a partner for 21 years at
Davis Gilbert, so welcome here.
And then on the other side of the
(01:57):
plane, we have, I would call it trial
services, but in terms of, and we don't
wanna get to the point where we actually
go to trial, but if you do go
to trial, Ina is the other part of
this dynamic duo.
Ina Schroer is your partner at Davis and
Gilbert, Boston University Law.
(02:20):
But prior to coming on board to Davis
and Gilbert, you were an assistant district attorney
in the Manhattan district attorney's office.
So we know how busy that office is.
And when it comes to protecting what we
wanna do, even if it doesn't wanna go
to trial, we really welcome you to the
party today.
What we would like to start out with
is AI.
(02:41):
So I've started to see in just client
contracts, because with our bigger clients, we tend
to indemnify our clients.
We stand in the way of lawsuits.
We're responsible for training our people, putting through
all of the knockout and the trademark claims
and getting everything done prior to even presenting
(03:03):
concepts to our clients.
But with AI and the generative side of
it, I'm gonna throw it to you first,
Ashima.
What are you seeing from a contractual point
of view or a watch out point of
view on generative AI, particularly when people in
the concept stage are using it for storyboard
creation, image creation, even content creation for blogs
(03:26):
and podcasts, for example.
What do you think about that?
So first, thanks for having us.
I appreciate it.
We're excited to be here and to talk
about our practice and what we do for
clients, including 500 Degrees.
I wanna take a half a step back
on AI.
AI has become a radioactive term and clients
reflexively are telling their agencies not to use
(03:48):
AI.
That's what we are seeing because they're responding
to issues that have arisen that they've heard
about, frankly, many of them have never experienced
anything that resembles a dispute, but they've heard
that AI is problematic and they don't distinguish
between the different types of AI.
And many of them are already using AI
in their own business and their agencies are
(04:09):
using it.
They just don't realize their agencies are using
it.
So there's lots of different kinds of AI
and it's incredibly effective, efficient, cost savings when
you're crunching massive amounts of data.
So when you're targeting ads, when you're buying
ads, if you're doing anything in the programmatic
space, if you are trying to create narrow
(04:30):
audiences for products, and I know in the
QSR space, creating very narrow target audiences is
incredibly important.
You're using AI, whether you realize it or
not.
So if a client suddenly says, you know
what, I've been hearing about AI.
I wanna make sure you guys aren't using
AI.
It's really important to hit pause and make
sure they understand what kind of AI you're
already using and what kind of AI they
(04:50):
think they're objecting to.
But you mentioned generative AI and that is
the kind that they really are objecting to,
that clients really are objecting to, whether they
know it or not.
So generative AI is a form of artificial
intelligence that when you give it commands, it
creates content based on what's been fed to
it.
So it's got a lot of human intervention,
(05:10):
whether people realize it or not, but that
human intervention largely exists on the training side.
So you're training the AI engine.
So you are feeding it stuff.
You're feeding it third-party content more of
the time, if not all of the time.
And then it's generating something.
It could be generating content that is copy.
It could be generating content that is an
(05:31):
image.
It could be generating content that is a
video.
And that's where clients are hitting pause or
even hitting a full red light stop button
because they don't want to have the problems
that they've heard about.
And those problems fall into a number of
places.
The first one, and this is kind of
the lightest concern they have is confidentiality.
Often you are asking the AI engine, which
(05:52):
you don't own.
It's a third-party AI engine and people
are very familiar with chat GPT, but certainly
there's many others.
And the questions you are asking it go
into its question bank and have no confidentiality.
So you can often tip off what your
client is doing.
If you're doing the Superbowl spot and you're
asking it to generate images of cats playing
the piano, that engine now knows who you
(06:15):
are and what kind of query you are
interested in.
And the image that it renders, part two,
is generally not something that they are giving
to you exclusively.
So there's non-exclusivity when your client may
have wanted, for example, to maintain confidentiality and
also to own what it's creating.
The third part, the part that they are
(06:36):
really alarmed about, the clients are alarmed about,
and rightly so, is that quite often what
that engine creates, what that AI engine creates,
is it's using third-party content and it's
doing so in a way that has no
permission associated with it.
So it very often could be deemed to
be infringing.
There's a lot of discussion that's been going
(06:58):
on in legal journals and in even the
popular press as to whether or not something
is infringing when you can't pick out the
original components, when you can't tell what was
fed into it, because the thing that is
rendered by the AI engine looks nothing like
it.
Has infringement really occurred, right?
If you feed it 600 photos of cats
and then you ask it to produce an
(07:18):
image of a cat and the cat photo
that comes out looks nothing like the 600
photos, but just has the general contours of
a cat, is it still infringing any of
those individual photos when you cannot find anything
that is even similar to any of those
original 600 in the mishmash that the AI
engine produces?
So there's lots of concerns with generative AI
(07:39):
and I've only touched upon some of them.
You can often protect yourself between you and
your client through careful contract terms, things like
indemnity, things like assurances of no originality, things
about when we use an AI engine, we
cannot assure you, the client, that you will
own exclusively all of what you produce.
(08:00):
That's not to say your client will sign
on to that, but you as an agency
can try to get in terms so that
you can use AI engines shifting any responsibility
onto the client or at least trying to
assure them of certain kinds of protections that
you can provide or shifting away from the
(08:22):
protections you can't provide.
Indemnity is a place where it's important.
Reps and warranties, making sure you're not giving
any reps and warranties of exclusivity, reps and
warranties of originality, reps and warranties of no
third-party materials were used, all of those
things can go into a contract.
That's not to say clients will sign on
to them, as I said, but there are
ways to make sure you are able to
(08:43):
use AI.
The other thing is to have a conversation
with your client about what AI is and
make sure they understand, as I said at
the beginning, because AI can be a huge
cost savings.
You know, we have clients right now that
at least for comps, not public facing material,
but at least for comps, and comps include
both regular, you know, still kind of mood
(09:03):
board comps and then full-on, you know,
animatics, if they're doing a television commercial or
something else in the video space, where they're
using AI in order to generate what they
need to give the client a look and
feel, an idea of what the final production
will look like.
And it is so much more effective because
it's quicker and less expensive to do it
(09:24):
that way.
So that's a sort of middle place where
you and clients can often meet in that
they will save a lot of money, they
will save a lot of time.
It is a lot easier on the agency.
It's just a hell of a lot less
work than, you know, sort of putting together
a mood board or a clip reel, a
sizzle reel, but just ensuring them that none
of this will be public facing.
(09:44):
There's lots of ways to use it effectively.
Well, I think, you know, and I'm gonna
pop over to Ina here in terms of
it's a new field.
So you probably haven't seen any trial yet,
but it's gonna come.
And it's either going to be a relationship
issue between agency and client in terms of
(10:06):
contract violation or whatever the terms that we've
agreed to based on the definitions of AI.
Because I see your point, Ashima, the idea
of discussing what the terms are and the
usage and the fair rights and where we
stop and where we start.
But I know on your side, what are
you gonna anticipate doing either between client and
(10:27):
agency in terms of disagreement or perhaps somebody
who's out there, a rapacious firm that says,
we found something that we think is leading
and improper and or false.
What do you see about the future of
protecting ourselves or protecting our clients?
(10:47):
So that's where it's actually pretty tricky to
kind of predict how the courts are gonna
view these issues.
And that's one of the things that's so
interesting about advertising law.
There's always something new that no one had
ever thought of before.
And we try to protect our clients by
anticipating what the legal risks could be.
(11:07):
But in an area like this, where the
technology is outpacing the law and the regulations,
it's sometimes hard to predict, but we use
past experience and analogous situations.
And what we really are trying to keep
an eye out for is there are concerns
about the regulators, but there are also greater
concerns about class action, plaintiff's lawyers who carefully
(11:31):
monitor the market are always keeping a very
close eye on the advertising industry and quickly
jump on to various issues and start making
claims and testing what the legal precedent will
turn out to be.
And I think that the area that's a
risk, that's an obvious risk with AI is
(11:52):
false or misleading communications.
That there's something inherently deceptive about the way
that images are being created, messages are being
conveyed through the use of AI.
And we don't yet have a clear way
to inform consumers and consumers may not yet
(12:13):
have an understanding of the significance of AI,
what it means.
And so it's hard to give the proper
disclosures and caveats to images and messages that
are being conveyed through advertising using this new
technology.
So, when you're in the process of mediating
or at least looking at the merits of
(12:35):
an argument that has been raised against a
client or an agency, and you're looking for
that evidence trail, the proof, would you suggest
it's a good idea to, I don't know
if it's an affidavit, I'm not sure what
the term would be, but to lock down
the steps of everything that we've done to
(12:58):
create an ad, an image, a piece of
digital display work, what have you.
How would you recommend to clients in the
agency world to create that trail?
So if we ever have to come back
and do it, it's presented well.
Very often there are cases that are brought
years after the content was created.
(13:22):
And very often there's been turnover in an
agency and the people who were directly involved
may no longer be involved, may no longer
be there and able to answer questions.
And so sometimes it is very difficult to
piece together the steps that were made.
And even agencies that have good record keeping
and good policies, those policies are only as
(13:44):
good as the compliance with them.
And when people are busy and trying to
meet deadlines and satisfy clients, they're not always
conscientious about keeping the record of what actually
occurred.
And so it's very important to keep that
in mind, but it's also important to keep
in mind that more casual conversations, particularly by
(14:06):
email and now more by text and by
Slack, people say things in those communications without
recognizing that those are discoverable and could be
required to be produced to the other side
if you end up in a litigation.
And so it's more often that the risks
(14:26):
and the challenges are presented in people very
casually making offhand comments that could later be
difficult to explain out of context in a
deposition or if you end up at trial.
And Ashima, back to you, I'm gonna follow
a thread here in terms of the actual
(14:47):
either graphic input or editorial input into a
generative platform.
If we count through just the testing, just
the people that are being casually interested in
how all of this works, but it invades
the workforce and somehow IP is, or a
(15:11):
brief or information or confidential information is pushed
into a generative platform to like, hey, explore,
I just got a brief from a client
and I throw it into chat GPT.
We're already, we've already trained our people to,
you know, that's verboten, you can't do that.
But do you see a bit of that
as a restraint or something that everybody has
(15:32):
to be on top of?
I think agencies shouldn't be, I mean, my
personal opinion is that agencies shouldn't be making
this verboten, they should make rules for how
to use it, but it's too efficient an
engine.
So the first thing I would say is
there's hallucinations, that's the term that's used for
basically the crap that an AI engine, the
(15:53):
chat GPT using them as an example, although
they're certainly not the only ones and I
don't mean to suggest that everything they produce
is inaccurate, but just the same way that
you feed in photos and the photo you
get back has three hands in it, that's
sort of the classic, right?
Is that the photo you get back, for
whatever reason, AI engines have a very difficult
time with limbs.
They seem to produce lots of photos that
(16:15):
have three hands or six fingers.
Or misspelling.
Or misspelling, so that's on the tech side,
right?
So they also produce false answers, so there's
been a lot of cases, there's one involving
Michael Cohen, Trump's, I guess, former attorney where
he produced a brief to a court, in
a court case, which I believe he was
the defendant in, that had just fabricated cases.
These cases did not exist, they looked exactly
(16:38):
like cases would look, but they were just,
there was no court, there was no decision,
there was no judge, it just made up
cases.
These same kind of hallucinations happen on the
factual side in advertising.
So if you're trying to produce, you're doing
a creative brief and you've got a new
chicken product and you want it to produce
information on, and it's, let's say it's organic,
(16:59):
and you want information on organic farming, do
not trust, so back to your question, do
not trust the information that an AI engine
produces.
It is the start of an investigation, it's
sort of the ground zero, but it's not
where you should land.
The second, and this has gotten a little
bit of attention, and I think it should
get more attention, is the fact that there
(17:19):
are enormous biases in what AI produces.
So if you ask it to produce a
picture of a professional woman, 99% of
the time, and I'm exaggerating by a huge
amount, but it will produce a picture of
a white woman, it will not produce a
picture of a woman of color.
If you ask it to produce a picture
of a woman of color, it will produce
(17:40):
a light-skinned woman of color.
So it's, and that's not because the AI
engine is, I'm gonna just say racist as
shorthand, it's because the images that have been
fed into it are images that show professional
white women because there's just fewer pictures out
there that are being fed, or just fewer
pictures out there full stop of women of
(18:00):
color carrying briefcases, wearing business suits.
So the AI engine is often producing images
that just perpetuate biases, and brands need to
be cognizant of that, need to be sensitive
of the fact that some of the information
that they're receiving out of an AI engine
may, they need to step back and take
a look at that and say, wait a
minute, is this really what we want?
(18:20):
And is this really the message that we
need to be putting out there?
And some of that has gotten attention, but
I think that that's something that people are,
as wokeness increases in the advertising industry, I
think that that's getting more attention.
One of the things that's interesting is in
Europe, in the last few weeks, they produced
this AI Act.
(18:42):
It covers all sort of 27 EU nations,
and it requires very careful use of AI,
but two of the things probably most relevant
to the ad industry is that it requires
disclosure on advertising.
You have to say when an ad was
produced using AI, that's not something we need
to do in the US outside of the
(19:02):
political space.
Political advertising needs to disclose when AI was
involved in certain states.
But the second one is that they also
have to indicate, they have to report, not
publicly, but to like a centralized board, what
information was used to generate the AI, how
they trained that AI engine, because that's information
(19:22):
that's being compiled as well.
I assume for bias reasons, but perhaps also
going back to Ina's point to make sure
there is a record of what was fed
in because memories change, people disappear.
It's gonna be hard if people didn't keep
good briefing books for the client work they
did to know what the source material was
for an ad.
So that's something that's happening in Europe that
(19:44):
we haven't yet adopted in the US, but
Biden's been very active in this space.
Everything that happens in Europe then goes to
California, then sweeps across the US.
Ashima, you mentioned politics, I believe.
We're entering a political race that is contentious
to say the least between two parties.
(20:05):
We've seen disinformation campaigns in the last election.
And not to say that each of the
parties or enemy government actors aren't beneath the
use of AI, but the ability for we
as voters to understand what is real, what
(20:26):
is not, I guess, trust your instinct and
be a smart buyer was certainly the first
course of action for all of us as
voters.
But how do we pick apart what's happening
in terms of either a Joe Biden ad
or Donald Trump ad that somebody has used
visual AI to create words and thoughts and
(20:49):
deeds that perhaps that candidate has never said
before in their life?
And it's in and it's out.
It seems on the political side, they get
a pass.
What's happening there in terms of the laws
that we already have on the books and
the enforcement of them?
I would not say that people are getting
a pass.
It may seem like they are getting a
(21:10):
pass, but there are challenges going on.
And I wanna just take a step back
when you said, we're not sure what happened
in that last election.
If one is to believe stories in the
Washington Post, New York Times and other mainstream
national newspapers, there was a decent amount of
reporting about foreign interference with subgroups in messaging
(21:32):
that was put out, especially in messaging that
was put out in connection with a lot
of female-centered organizations where messages were put
out using AI, targeting individuals, trying to send
positive or negative messages about candidates that were
just entirely fabricated.
(21:52):
So that, again, the sort of what we
talked about AI and how good it is
at micro-targeting because it can crunch vast
amounts of data and identify audiences that was
used in the political space.
The same way it could try to sell
you a cable subscription or sell you a
new car because it knows you need a
minivan.
They were also able to target individuals based
(22:12):
on the very specific interests that they had.
If you're interested in immigration, it had a
message often false about one of the candidates
immigration platform.
There are false advertising laws on the books
in every state and very rarely do candidates
make claims against each other but they do
(22:34):
on occasion happen.
They happen more on the lower level races,
statewide races, not on the national races, but
they do occasionally happen.
The difficulty in bringing those cases is that
there's obviously First Amendment protections for political speech
that doesn't exist for commercial speech.
(22:56):
Quite often, it's not about monetary damages.
It's about stopping the ad and by the
time the ad has run, right?
Because if I'm in the Biden campaign, I'm
not gonna sue the Trump campaign for money.
It's not money I want.
I want them to stop running the ad.
And the ad has run for such a
short period of time already.
It's off the air by the time the
damage is done, right?
You can't sort of unspin that.
(23:18):
Or as Ina often likes to say, you
can't, is it unring the bell you like
to say or can't put the horse back
in the barn?
I forget which one is the phrase.
Both.
I mean, Ina, you just wouldn't have time
to be able to bring something to court.
By then it's done and the campaign's over.
So do you think people rely on that?
(23:40):
I mean, there are some emergency applications.
You can seek injunctive relief to get something
addressed more quickly.
But very often, clients come to us and
say, we need to remedy this.
We need to do something quickly.
And often they're disappointed to learn that litigation
and going to trial is a very slow,
(24:03):
expensive and burdensome process that isn't gonna get
them the quick result that they're hoping for.
And so I think in the political arena,
we're relying on the press to fact check
things and call out things that are blatantly
false or misinformation.
And they are going to do that with
images as well.
(24:24):
You know, if something has been altered in
a visual representation, now people are aware of
that.
And I think that that's gonna be more
of an issue and certainly is something that
we're seeing more of in terms of monitoring
and addressing in public forums, if not in
litigation.
And one of the problems is that the
(24:44):
FEC has a regulation, at least in the
television and radio space, they have a regulation
that broadcasters are immunized for the content, that
political content they run, as long as they
do not alter it.
So if you see an ad, if you're
a television station, television network, you cannot change
that ad.
(25:05):
You are not allowed to touch it in
any way.
You are obliged to run that.
In exchange for that, you get immunity for
that ad.
And do they clear it?
Do they have to clear storyboards or do
preclearance on any of their advertising for politics?
I shouldn't know this because I'm an advertising
person, but I don't.
That's legal team.
Or, and, or you don't have any at
yet, you don't have any political candidates.
(25:25):
So you would learn it when you need
to learn it.
But no, that doesn't happen.
They in fact required, there's a lowest unit
price, they're required to sell it, sell that
ad time at the lowest price that they
have to sell that at the same price
to all qualified candidates, all sorts of things
that the media makes a lot of money
during the election season, but they also don't
(25:47):
get to charge the highest unit price, et
cetera, et cetera.
There's all sorts of constraints on the media.
They also get a free pass in the
way they wouldn't for a false QSR ad.
They get a free pass for a false
political ad in exchange for not making changes
to that ad.
Political advertising is fascinating and people, if they're
sort of sitting on the sidelines watching it,
(26:09):
especially in the AI space, it's gonna be
right now in the States, not the federal
level that they're gonna see regulation.
The States are the ones who are saying
that you cannot run an ad that uses
AI without disclosing it uses AI.
And you used an example, Matt, that I
think is a great example because it's an
actual real world example, not a hypothetical, where
(26:30):
people are animating the mouths of candidates and
putting words in their mouths that they never
said.
And from a viewer perspective, I mean, this
is the best damn lip syncing you're going
to see.
I saw one the other day and I
swear the candidate, the mouth aligned perfectly with
the words they were saying.
Yeah, it's not like clutch cargo and that
(26:51):
famous cartoon where it's animated except for the
lips or some live reader in the background.
And you only had to feed in six
words for them to get the vocal tone
right.
You've got like three speeches, you feed those
in and they can do a full made
up, all of a sudden, Trump is supporting
immigration and Biden is whatever, Trumpian platform he
(27:15):
might otherwise reject.
It's astonishing how good this stuff is.
And you can see why the EU is
saying you have to disclose when you use
AI and the same thing with the States
because it's really frightening because people aren't going
to read the disclosure.
You know, they need to, in some ways
they need to outlaw it.
We could talk for a long time on
(27:36):
all these issues and this will not be
your last appearance on Burning Questions, but I
have one last Burning Question for each of
you and I'll start with Ina.
What are you reading?
What's on your bedside table that you're reading
that you're fascinated with or what's on your
playlist for music?
(27:56):
And talk to us, what are you interested?
Even podcasts you listen to besides Burning Questions.
I am a huge reader and so I
am always reading something.
I skew toward nonfiction and I just finished
Shoe Dog, which is Phil Knight's memoir about
(28:18):
the founding of Nike.
And it is a fascinating read.
He's a great writer and a very inspirational
story.
So I highly recommend it.
The Ben Affleck film associated with Shoe Dog
was incredible.
I thought that was great.
So I'll pick that up and we're going
to put that out.
We're going to put that out for our
(28:38):
listeners for like the top ad legends reading
list.
Now, Ashima, before we close, what are you
reading or what are you listening to?
What podcasts fascinate you?
Okay, so what I'm listening to is Maggie
Rogers' new album.
I'm obsessed with Maggie Rogers.
I'm like three times the age of everyone
who listens to Maggie Rogers, but I am
(29:00):
absolutely obsessed with her.
And I have listened to Beyonce's new album.
I need to listen to it more.
I'm a huge Jolene fan.
I'm trying to decide if I'm okay with
her changing all of the lyrics.
I understand Dolly Parton has fully endorsed it,
but I'm not quite there yet.
I think I need to listen to it
a little bit more.
What I've heard I like, and I really
(29:21):
do like Beyonce.
And then what I'm reading, normally I read
only fiction.
I love to escape in fiction, but I
read an essay by this feminist writer named
Rebecca Solnit recently.
And someone recommended to me the essay.
And then I picked up a book of
her essays.
Her book has the most wonderful name.
It's Men Explain Things to Me.
(29:43):
And it's fantastic that the essay that was
given to me is called Men Explain Things
to Me.
And it's about how she was at a
party and someone asked her what she does.
She said, she's a writer.
And they asked her what she was writing
about.
And she told them what she had written
about.
(Transcribed by TurboScribe.ai. Go Unlimited to remove this message.) Welcome to Burning Questions from 500 Degrees.
We work together as a company, 500 Degrees
is a content fulfillment shop.
We have offices in Toronto, Miami, Columbus, Ohio,
and we work with some of the best
QSRs in the country, convenience stores and new
(00:31):
clients in the automotive industry, so we're expanding.
And we create thousands of images and assets
that are deployed on multiple digital platforms.
And if you wanna learn more about us,
go to weare500degrees.com and give us a
shout and we'll get back to you.
The best thing about my job is the
(00:52):
people that I come into contact during our
work day.
And it's great vendors, great associates, great people,
great law firms.
And that's what we wanna share in this
podcast.
And today we welcome the folks from Davis
and Gilbert.
I'm gonna call you guys the dynamic duo.
(01:14):
I've been working with Davis Gilbert for over
25 years and you guys have always been
great stewards of our legal issues and keeping
us out of trouble and just keeping us
on the path from contracts to trademarks to
whatever it is that we have to do
from claims.
I'm gonna first introduce Ashima Dayal.
(01:35):
And Ashima, you work on the counsel side,
I would call it, in terms of trademark
clearance, claims clearance, IP protection, numerous activities on
your side, Columbia Law and a background in
art history.
You've been a partner for 21 years at
Davis Gilbert, so welcome here.
And then on the other side of the
(01:57):
plane, we have, I would call it trial
services, but in terms of, and we don't
wanna get to the point where we actually
go to trial, but if you do go
to trial, Ina is the other part of
this dynamic duo.
Ina Schroer is your partner at Davis and
Gilbert, Boston University Law.
(02:20):
But prior to coming on board to Davis
and Gilbert, you were an assistant district attorney
in the Manhattan district attorney's office.
So we know how busy that office is.
And when it comes to protecting what we
wanna do, even if it doesn't wanna go
to trial, we really welcome you to the
party today.
What we would like to start out with
is AI.
(02:41):
So I've started to see in just client
contracts, because with our bigger clients, we tend
to indemnify our clients.
We stand in the way of lawsuits.
We're responsible for training our people, putting through
all of the knockout and the trademark claims
and getting everything done prior to even presenting
(03:03):
concepts to our clients.
But with AI and the generative side of
it, I'm gonna throw it to you first,
Ashima.
What are you seeing from a contractual point
of view or a watch out point of
view on generative AI, particularly when people in
the concept stage are using it for storyboard
creation, image creation, even content creation for blogs
(03:26):
and podcasts, for example.
What do you think about that?
So first, thanks for having us.
I appreciate it.
We're excited to be here and to talk
about our practice and what we do for
clients, including 500 Degrees.
I wanna take a half a step back
on AI.
AI has become a radioactive term and clients
reflexively are telling their agencies not to use
(03:48):
AI.
That's what we are seeing because they're responding
to issues that have arisen that they've heard
about, frankly, many of them have never experienced
anything that resembles a dispute, but they've heard
that AI is problematic and they don't distinguish
between the different types of AI.
And many of them are already using AI
in their own business and their agencies are
(04:09):
using it.
They just don't realize their agencies are using
it.
So there's lots of different kinds of AI
and it's incredibly effective, efficient, cost savings when
you're crunching massive amounts of data.
So when you're targeting ads, when you're buying
ads, if you're doing anything in the programmatic
space, if you are trying to create narrow
(04:30):
audiences for products, and I know in the
QSR space, creating very narrow target audiences is
incredibly important.
You're using AI, whether you realize it or
not.
So if a client suddenly says, you know
what, I've been hearing about AI.
I wanna make sure you guys aren't using
AI.
It's really important to hit pause and make
sure they understand what kind of AI you're
already using and what kind of AI they
(04:50):
think they're objecting to.
But you mentioned generative AI and that is
the kind that they really are objecting to,
that clients really are objecting to, whether they
know it or not.
So generative AI is a form of artificial
intelligence that when you give it commands, it
creates content based on what's been fed to
it.
So it's got a lot of human intervention,
(05:10):
whether people realize it or not, but that
human intervention largely exists on the training side.
So you're training the AI engine.
So you are feeding it stuff.
You're feeding it third-party content more of
the time, if not all of the time.
And then it's generating something.
It could be generating content that is copy.
It could be generating content that is an
(05:31):
image.
It could be generating content that is a
video.
And that's where clients are hitting pause or
even hitting a full red light stop button
because they don't want to have the problems
that they've heard about.
And those problems fall into a number of
places.
The first one, and this is kind of
the lightest concern they have is confidentiality.
Often you are asking the AI engine, which
(05:52):
you don't own.
It's a third-party AI engine and people
are very familiar with chat GPT, but certainly
there's many others.
And the questions you are asking it go
into its question bank and have no confidentiality.
So you can often tip off what your
client is doing.
If you're doing the Superbowl spot and you're
asking it to generate images of cats playing
the piano, that engine now knows who you
(06:15):
are and what kind of query you are
interested in.
And the image that it renders, part two,
is generally not something that they are giving
to you exclusively.
So there's non-exclusivity when your client may
have wanted, for example, to maintain confidentiality and
also to own what it's creating.
The third part, the part that they are
(06:36):
really alarmed about, the clients are alarmed about,
and rightly so, is that quite often what
that engine creates, what that AI engine creates,
is it's using third-party content and it's
doing so in a way that has no
permission associated with it.
So it very often could be deemed to
be infringing.
There's a lot of discussion that's been going
(06:58):
on in legal journals and in even the
popular press as to whether or not something
is infringing when you can't pick out the
original components, when you can't tell what was
fed into it, because the thing that is
rendered by the AI engine looks nothing like
it.
Has infringement really occurred, right?
If you feed it 600 photos of cats
and then you ask it to produce an
(07:18):
image of a cat and the cat photo
that comes out looks nothing like the 600
photos, but just has the general contours of
a cat, is it still infringing any of
those individual photos when you cannot find anything
that is even similar to any of those
original 600 in the mishmash that the AI
engine produces?
So there's lots of concerns with generative AI
(07:39):
and I've only touched upon some of them.
You can often protect yourself between you and
your client through careful contract terms, things like
indemnity, things like assurances of no originality, things
about when we use an AI engine, we
cannot assure you, the client, that you will
own exclusively all of what you produce.
(08:00):
That's not to say your client will sign
on to that, but you as an agency
can try to get in terms so that
you can use AI engines shifting any responsibility
onto the client or at least trying to
assure them of certain kinds of protections that
you can provide or shifting away from the
(08:22):
protections you can't provide.
Indemnity is a place where it's important.
Reps and warranties, making sure you're not giving
any reps and warranties of exclusivity, reps and
warranties of originality, reps and warranties of no
third-party materials were used, all of those
things can go into a contract.
That's not to say clients will sign on
to them, as I said, but there are
ways to make sure you are able to
(08:43):
use AI.
The other thing is to have a conversation
with your client about what AI is and
make sure they understand, as I said at
the beginning, because AI can be a huge
cost savings.
You know, we have clients right now that
at least for comps, not public facing material,
but at least for comps, and comps include
both regular, you know, still kind of mood
(09:03):
board comps and then full-on, you know,
animatics, if they're doing a television commercial or
something else in the video space, where they're
using AI in order to generate what they
need to give the client a look and
feel, an idea of what the final production
will look like.
And it is so much more effective because
it's quicker and less expensive to do it
(09:24):
that way.
So that's a sort of middle place where
you and clients can often meet in that
they will save a lot of money, they
will save a lot of time.
It is a lot easier on the agency.
It's just a hell of a lot less
work than, you know, sort of putting together
a mood board or a clip reel, a
sizzle reel, but just ensuring them that none
of this will be public facing.
(09:44):
There's lots of ways to use it effectively.
Well, I think, you know, and I'm gonna
pop over to Ina here in terms of
it's a new field.
So you probably haven't seen any trial yet,
but it's gonna come.
And it's either going to be a relationship
issue between agency and client in terms of
(10:06):
contract violation or whatever the terms that we've
agreed to based on the definitions of AI.
Because I see your point, Ashima, the idea
of discussing what the terms are and the
usage and the fair rights and where we
stop and where we start.
But I know on your side, what are
you gonna anticipate doing either between client and
(10:27):
agency in terms of disagreement or perhaps somebody
who's out there, a rapacious firm that says,
we found something that we think is leading
and improper and or false.
What do you see about the future of
protecting ourselves or protecting our clients?
(10:47):
So that's where it's actually pretty tricky to
kind of predict how the courts are gonna
view these issues.
And that's one of the things that's so
interesting about advertising law.
There's always something new that no one had
ever thought of before.
And we try to protect our clients by
anticipating what the legal risks could be.
(11:07):
But in an area like this, where the
technology is outpacing the law and the regulations,
it's sometimes hard to predict, but we use
past experience and analogous situations.
And what we really are trying to keep
an eye out for is there are concerns
about the regulators, but there are also greater
concerns about class action, plaintiff's lawyers who carefully
(11:31):
monitor the market are always keeping a very
close eye on the advertising industry and quickly
jump on to various issues and start making
claims and testing what the legal precedent will
turn out to be.
And I think that the area that's a
risk, that's an obvious risk with AI is
(11:52):
false or misleading communications.
That there's something inherently deceptive about the way
that images are being created, messages are being
conveyed through the use of AI.
And we don't yet have a clear way
to inform consumers and consumers may not yet
(12:13):
have an understanding of the significance of AI,
what it means.
And so it's hard to give the proper
disclosures and caveats to images and messages that
are being conveyed through advertising using this new
technology.
So, when you're in the process of mediating
or at least looking at the merits of
(12:35):
an argument that has been raised against a
client or an agency, and you're looking for
that evidence trail, the proof, would you suggest
it's a good idea to, I don't know
if it's an affidavit, I'm not sure what
the term would be, but to lock down
the steps of everything that we've done to
(12:58):
create an ad, an image, a piece of
digital display work, what have you.
How would you recommend to clients in the
agency world to create that trail?
So if we ever have to come back
and do it, it's presented well.
Very often there are cases that are brought
years after the content was created.
(13:22):
And very often there's been turnover in an
agency and the people who were directly involved
may no longer be involved, may no longer
be there and able to answer questions.
And so sometimes it is very difficult to
piece together the steps that were made.
And even agencies that have good record keeping
and good policies, those policies are only as
(13:44):
good as the compliance with them.
And when people are busy and trying to
meet deadlines and satisfy clients, they're not always
conscientious about keeping the record of what actually
occurred.
And so it's very important to keep that
in mind, but it's also important to keep
in mind that more casual conversations, particularly by
(14:06):
email and now more by text and by
Slack, people say things in those communications without
recognizing that those are discoverable and could be
required to be produced to the other side
if you end up in a litigation.
And so it's more often that the risks
(14:26):
and the challenges are presented in people very
casually making offhand comments that could later be
difficult to explain out of context in a
deposition or if you end up at trial.
And Ashima, back to you, I'm gonna follow
a thread here in terms of the actual
(14:47):
either graphic input or editorial input into a
generative platform.
If we count through just the testing, just
the people that are being casually interested in
how all of this works, but it invades
the workforce and somehow IP is, or a
(15:11):
brief or information or confidential information is pushed
into a generative platform to like, hey, explore,
I just got a brief from a client
and I throw it into chat GPT.
We're already, we've already trained our people to,
you know, that's verboten, you can't do that.
But do you see a bit of that
as a restraint or something that everybody has
(15:32):
to be on top of?
I think agencies shouldn't be, I mean, my
personal opinion is that agencies shouldn't be making
this verboten, they should make rules for how
to use it, but it's too efficient an
engine.
So the first thing I would say is
there's hallucinations, that's the term that's used for
basically the crap that an AI engine, the
(15:53):
chat GPT using them as an example, although
they're certainly not the only ones and I
don't mean to suggest that everything they produce
is inaccurate, but just the same way that
you feed in photos and the photo you
get back has three hands in it, that's
sort of the classic, right?
Is that the photo you get back, for
whatever reason, AI engines have a very difficult
time with limbs.
They seem to produce lots of photos that
(16:15):
have three hands or six fingers.
Or misspelling.
Or misspelling, so that's on the tech side,
right?
So they also produce false answers, so there's
been a lot of cases, there's one involving
Michael Cohen, Trump's, I guess, former attorney where
he produced a brief to a court, in
a court case, which I believe he was
the defendant in, that had just fabricated cases.
These cases did not exist, they looked exactly
(16:38):
like cases would look, but they were just,
there was no court, there was no decision,
there was no judge, it just made up
cases.
These same kind of hallucinations happen on the
factual side in advertising.
So if you're trying to produce, you're doing
a creative brief and you've got a new
chicken product and you want it to produce
information on, and it's, let's say it's organic,
(16:59):
and you want information on organic farming, do
not trust, so back to your question, do
not trust the information that an AI engine
produces.
It is the start of an investigation, it's
sort of the ground zero, but it's not
where you should land.
The second, and this has gotten a little
bit of attention, and I think it should
get more attention, is the fact that there
(17:19):
are enormous biases in what AI produces.
So if you ask it to produce a
picture of a professional woman, 99% of
the time, and I'm exaggerating by a huge
amount, but it will produce a picture of
a white woman, it will not produce a
picture of a woman of color.
If you ask it to produce a picture
of a woman of color, it will produce
(17:40):
a light-skinned woman of color.
So it's, and that's not because the AI
engine is, I'm gonna just say racist as
shorthand, it's because the images that have been
fed into it are images that show professional
white women because there's just fewer pictures out
there that are being fed, or just fewer
pictures out there full stop of women of
(18:00):
color carrying briefcases, wearing business suits.
So the AI engine is often producing images
that just perpetuate biases, and brands need to
be cognizant of that, need to be sensitive
of the fact that some of the information
that they're receiving out of an AI engine
may, they need to step back and take
a look at that and say, wait a
minute, is this really what we want?
(18:20):
And is this really the message that we
need to be putting out there?
And some of that has gotten attention, but
I think that that's something that people are,
as wokeness increases in the advertising industry, I
think that that's getting more attention.
One of the things that's interesting is in
Europe, in the last few weeks, they produced
this AI Act.
(18:42):
It covers all sort of 27 EU nations,
and it requires very careful use of AI,
but two of the things probably most relevant
to the ad industry is that it requires
disclosure on advertising.
You have to say when an ad was
produced using AI, that's not something we need
to do in the US outside of the
(19:02):
political space.
Political advertising needs to disclose when AI was
involved in certain states.
But the second one is that they also
have to indicate, they have to report, not
publicly, but to like a centralized board, what
information was used to generate the AI, how
they trained that AI engine, because that's information
(19:22):
that's being compiled as well.
I assume for bias reasons, but perhaps also
going back to Ina's point to make sure
there is a record of what was fed
in because memories change, people disappear.
It's gonna be hard if people didn't keep
good briefing books for the client work they
did to know what the source material was
for an ad.
So that's something that's happening in Europe that
(19:44):
we haven't yet adopted in the US, but
Biden's been very active in this space.
Everything that happens in Europe then goes to
California, then sweeps across the US.
Ashima, you mentioned politics, I believe.
We're entering a political race that is contentious
to say the least between two parties.
(20:05):
We've seen disinformation campaigns in the last election.
And not to say that each of the
parties or enemy government actors aren't beneath the
use of AI, but the ability for we
as voters to understand what is real, what
(20:26):
is not, I guess, trust your instinct and
be a smart buyer was certainly the first
course of action for all of us as
voters.
But how do we pick apart what's happening
in terms of either a Joe Biden ad
or Donald Trump ad that somebody has used
visual AI to create words and thoughts and
(20:49):
deeds that perhaps that candidate has never said
before in their life?
And it's in and it's out.
It seems on the political side, they get
a pass.
What's happening there in terms of the laws
that we already have on the books and
the enforcement of them?
I would not say that people are getting
a pass.
It may seem like they are getting a
(21:10):
pass, but there are challenges going on.
And I wanna just take a step back
when you said, we're not sure what happened
in that last election.
If one is to believe stories in the
Washington Post, New York Times and other mainstream
national newspapers, there was a decent amount of
reporting about foreign interference with subgroups in messaging
(21:32):
that was put out, especially in messaging that
was put out in connection with a lot
of female-centered organizations where messages were put
out using AI, targeting individuals, trying to send
positive or negative messages about candidates that were
just entirely fabricated.
(21:52):
So that, again, the sort of what we
talked about AI and how good it is
at micro-targeting because it can crunch vast
amounts of data and identify audiences that was
used in the political space.
The same way it could try to sell
you a cable subscription or sell you a
new car because it knows you need a
minivan.
They were also able to target individuals based
(22:12):
on the very specific interests that they had.
If you're interested in immigration, it had a
message often false about one of the candidates
immigration platform.
There are false advertising laws on the books
in every state and very rarely do candidates
make claims against each other but they do
(22:34):
on occasion happen.
They happen more on the lower level races,
statewide races, not on the national races, but
they do occasionally happen.
The difficulty in bringing those cases is that
there's obviously First Amendment protections for political speech
that doesn't exist for commercial speech.
(22:56):
Quite often, it's not about monetary damages.
It's about stopping the ad and by the
time the ad has run, right?
Because if I'm in the Biden campaign, I'm
not gonna sue the Trump campaign for money.
It's not money I want.
I want them to stop running the ad.
And the ad has run for such a
short period of time already.
It's off the air by the time the
damage is done, right?
You can't sort of unspin that.
(23:18):
Or as Ina often likes to say, you
can't, is it unring the bell you like
to say or can't put the horse back
in the barn?
I forget which one is the phrase.
Both.
I mean, Ina, you just wouldn't have time
to be able to bring something to court.
By then it's done and the campaign's over.
So do you think people rely on that?
(23:40):
I mean, there are some emergency applications.
You can seek injunctive relief to get something
addressed more quickly.
But very often, clients come to us and
say, we need to remedy this.
We need to do something quickly.
And often they're disappointed to learn that litigation
and going to trial is a very slow,
(24:03):
expensive and burdensome process that isn't gonna get
them the quick result that they're hoping for.
And so I think in the political arena,
we're relying on the press to fact check
things and call out things that are blatantly
false or misinformation.
And they are going to do that with
images as well.
(24:24):
You know, if something has been altered in
a visual representation, now people are aware of
that.
And I think that that's gonna be more
of an issue and certainly is something that
we're seeing more of in terms of monitoring
and addressing in public forums, if not in
litigation.
And one of the problems is that the
(24:44):
FEC has a regulation, at least in the
television and radio space, they have a regulation
that broadcasters are immunized for the content, that
political content they run, as long as they
do not alter it.
So if you see an ad, if you're
a television station, television network, you cannot change
that ad.
(25:05):
You are not allowed to touch it in
any way.
You are obliged to run that.
In exchange for that, you get immunity for
that ad.
And do they clear it?
Do they have to clear storyboards or do
preclearance on any of their advertising for politics?
I shouldn't know this because I'm an advertising
person, but I don't.
That's legal team.
Or, and, or you don't have any at
yet, you don't have any political candidates.
(25:25):
So you would learn it when you need
to learn it.
But no, that doesn't happen.
They in fact required, there's a lowest unit
price, they're required to sell it, sell that
ad time at the lowest price that they
have to sell that at the same price
to all qualified candidates, all sorts of things
that the media makes a lot of money
during the election season, but they also don't
(25:47):
get to charge the highest unit price, et
cetera, et cetera.
There's all sorts of constraints on the media.
They also get a free pass in the
way they wouldn't for a false QSR ad.
They get a free pass for a false
political ad in exchange for not making changes
to that ad.
Political advertising is fascinating and people, if they're
sort of sitting on the sidelines watching it,
(26:09):
especially in the AI space, it's gonna be
right now in the States, not the federal
level that they're gonna see regulation.
The States are the ones who are saying
that you cannot run an ad that uses
AI without disclosing it uses AI.
And you used an example, Matt, that I
think is a great example because it's an
actual real world example, not a hypothetical, where
(26:30):
people are animating the mouths of candidates and
putting words in their mouths that they never
said.
And from a viewer perspective, I mean, this
is the best damn lip syncing you're going
to see.
I saw one the other day and I
swear the candidate, the mouth aligned perfectly with
the words they were saying.
Yeah, it's not like clutch cargo and that
(26:51):
famous cartoon where it's animated except for the
lips or some live reader in the background.
And you only had to feed in six
words for them to get the vocal tone
right.
You've got like three speeches, you feed those
in and they can do a full made
up, all of a sudden, Trump is supporting
immigration and Biden is whatever, Trumpian platform he
(27:15):
might otherwise reject.
It's astonishing how good this stuff is.
And you can see why the EU is
saying you have to disclose when you use
AI and the same thing with the States
because it's really frightening because people aren't going
to read the disclosure.
You know, they need to, in some ways
they need to outlaw it.
We could talk for a long time on
(27:36):
all these issues and this will not be
your last appearance on Burning Questions, but I
have one last Burning Question for each of
you and I'll start with Ina.
What are you reading?
What's on your bedside table that you're reading
that you're fascinated with or what's on your
playlist for music?
(27:56):
And talk to us, what are you interested?
Even podcasts you listen to besides Burning Questions.
I am a huge reader and so I
am always reading something.
I skew toward nonfiction and I just finished
Shoe Dog, which is Phil Knight's memoir about
(28:18):
the founding of Nike.
And it is a fascinating read.
He's a great writer and a very inspirational
story.
So I highly recommend it.
The Ben Affleck film associated with Shoe Dog
was incredible.
I thought that was great.
So I'll pick that up and we're going
to put that out.
We're going to put that out for our
(28:38):
listeners for like the top ad legends reading
list.
Now, Ashima, before we close, what are you
reading or what are you listening to?
What podcasts fascinate you?
Okay, so what I'm listening to is Maggie
Rogers' new album.
I'm obsessed with Maggie Rogers.
I'm like three times the age of everyone
who listens to Maggie Rogers, but I am
(29:00):
absolutely obsessed with her.
And I have listened to Beyonce's new album.
I need to listen to it more.
I'm a huge Jolene fan.
I'm trying to decide if I'm okay with
her changing all of the lyrics.
I understand Dolly Parton has fully endorsed it,
but I'm not quite there yet.
I think I need to listen to it
a little bit more.
What I've heard I like, and I really
(29:21):
do like Beyonce.
And then what I'm reading, normally I read
only fiction.
I love to escape in fiction, but I
read an essay by this feminist writer named
Rebecca Solnit recently.
And someone recommended to me the essay.
And then I picked up a book of
her essays.
Her book has the most wonderful name.
It's Men Explain Things to Me.
(29:43):
And it's fantastic that the essay that was
given to me is called Men Explain Things
to Me.
And it's about how she was at a
party and someone asked her what she does.
She said, she's a writer.
And they asked her what she was writing
about.
And she told them what she had written
about.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.