Welcome to The MadTech Podcast from ExchangeWire,
bringing you the latest news and views from the world of ad tech, media and marketing.
Hello and welcome to The MadTech Podcast.
I'm Grainne Reid, Head of Marketing at ExchangeWire.
And this week I'm delighted to be joined by Josh Hill, Senior Director, Head of Optimization EMEA at Kepler.
How are you doing, Josh?
Doing well.
Thank you so much for having me.
I'm really excited to be here.
Great to have you in the office.
Yeah.
I am joined.
We've got John Still, Head of Content at ExchangeWire.
John, how are you?
Great.
Well, this week we will be discussing OpenAI's hack on Hugging Face, the EU fine to Google for 890 million euro over certain apps or breaches,
as well as the judge halting Paramount-Warner Bros deal.
The first story, OpenAI reveals that during a test,
an autonomous AI agent escaped its testing environment and hacked into AI startup Hugging Face,
an incident OpenAI called "unprecedented", involving "state of the art cyber capabilities".
The breach has intensified debate over AI safety, autonomy, and the need for stronger regulatory guardrails.
Tongue twisters all around there.
Josh, I'm to come to you first on this story
and what do you think we need
to try and seriously consider these guardrails and the AI standards?
Yeah.
There certainly needs to be some level of regulation here and some independent third party accountability mechanisms as well.
Right. We're seeing in this instance OpenAI, but some of these other LLMs move from this sort of theoretical risk to cybersecurity to very tangible, autonomous, like offensive threats.
So yeah, I think we're at an inflection point where
there needs to be some sort of regulatory intervention to ensure there are standards that are being maintained and upheld.
Yeah, John, what are your thoughts on this too?
Oh, yeah, like we needed regulation yesterday or the week before or probably in early 2023 when this started really coming out.
I think it's easy sometimes to look at these proclamations of like our AI escaped,
look at what it can do as a bit of a tease for the market.
Sometimes it's like, we are this like our machine is this powerful, maybe invest a few more billions in us and what we can do.
But also the risks are real.
Like if, you know, this happened and
an AI from a lab broke its containment and found its way into another AI system, then yeah,
the time for guardrails and standards and third party verification and bodies for this is absolutely now.
Yeah.
Like I say, if not two years ago. People familiar with this podcast will know that I'm always calling for regulation across the board.
But it's dangerous stuff and it's, you know, this...
Hugging Face found this without knowing where it had come from and then OpenAI confessed to it and did a blog post about what had happened
and how it happened and how they were just, you know, how this thing had broken its containment.
But that's, this is sort of like the first case of this happening and it's frightening that of course it's possible.
It's literally artificial intelligence.
It will be able to do these things.
But it just doubles down, I think, if not on
the value and the possibilities of AI, then the need for us to regulate it as quickly as possible.
Yeah.
The CEO for Hugging Face calls for radical transparency.
What do you think radical transparency could look like in practice?
Yeah.
I think it's it's always a challenge when you have these privately held companies and really finding the level that's appropriate to let them sort of operate and innovate,
but also hold them to a high standard.
So really setting up these bodies that can do independent, like real third party audits, access to sort of key systems and ensuring there are
paper trails with within these companies like OpenAI that really documents all of these incidents.
I'm sure there are probably more that haven't been disclosed, right?
Like this is this is just one very public example right now.
But being able to hold them accountable with, you know, potential fines where necessary,
you know. This escaped a virtual sandbox environment,
but maybe there should be regulation about the hardware that's actually being used for these test environments
to ensure that these these models aren't able to break out into into other systems.
But there needs to be something systematic, traceable, verifiable that will really hold them to a a higher standard.
Yeah, I'm like kind of talking about the platforms and just general systems.
John, how do you think companies could rethink their approach then to this sort of security and internal testing?
I think we've ended up in this AI arms race basically between all the companies and we've been hearing for years that, oh no, don't regulate ourselves, else our international
competitors will get ahead of us and we'll lose the arms race and China will win and the Chinese models will win.
Don't slow us down guys, don't slow us down.
And that's never been sustainable, kind of, the accelerationist
element of it has always been concerning.
It is just trying to do what they can as quickly as possible,
and that's when guardrails get put aside
and that's when safety protocols get put aside for in the name of progress
and it ends up in dangerous times, I think.
And so they need to slow down and think about the guardrails, think about the securities needed.
And certainly, as Josh said, I think this like
third party access to the systems to be able to verify, to be able to ensure that there is protocols going on, that this is being managed.
It should have been, it should really be like route one stuff now, I think, in terms of what the possibilities of AI are.
Yeah, yeah.
And Josh, how do you think the impact of AI, what this will mean then for the future of media buying?
Yeah, there are huge implications for this.
Of course we've been talking about, you know, agentic buying and bidding and creative generation and all those sort of core processes to, you know, buying media.
But I think the real risk is there's already so much ad fraud.
Think about the sheer volume of, you know, currency trading
trading hands in a given day or week, the more money being spent, the the more rife that industry is with fraud or at least fraudulent actors.
So if we as advertisers are willing to cede more control to these LLMs to do all of these different points of the buying phase
and really like outsourcing some of these human elements to to these models,
that just provides an additional channel for fraud and for risk. We're seeding that control.
And if we don't have real clarity and transparency on how these models are acting on advertisers' behalf or on behalf of those fraudulent actors, then
we find ourselves in a position where we're putting, you know, trillions of pounds globally at risk.
Yep.
Yeah, I guess just kind of I'm sure you're going to have similar opinions there to Josh on kind of the risk management side.
But is there anything else that you want to add, John, to what this means for the future of media buying?
Yeah, I think it really does accentuate the need for human oversight for this stuff.
Like agentic AI and agentic media buying is obviously the future in the same way that programmatic was.
It's just something that is going to make the system easier, more flexible.
But with that comes everything that Josh said.
It will absolutely exacerbate the problem with ad fraud.
If you give a machine your credit card, it's gonna go and buy mad stuff.
We need the guardrails on it.
If you're leaving your media buying to an agentic system, it has to have really smart, really tight rules as to what it's buying, where it's going, else it could put your entire
year's ad spend in MFAs or fraudulent sites that will never get a click, never get a view, never get anything off it.
It will have to be, I think I'm just defending the role of the human media planner here.
I really want them to still
be able to use these tools and for it to be efficient and for it to be a good way of doing things for on all sides.
But it needs it still needs that smart direction before it makes any transaction or is just given the keys to the bank account, I think.
Yeah.
Okay, well, great.
I'm going to move us on now to our second story. The European Commission has fined Google a combined 890 million euro,
which is the equivalent of 760 million pounds for two separate breaches of the Digital Marketers Act, marking the first time
the tech giant has been penalized under the bloc's landmark gatekeeper rules.
Brussels found that Google gave preferential treatment to its own services over rivals' offerings within Google search,
resulting in a 460 million euro fine and a further 430 million euro penalty was issued over Google Play, where the company was found to have
restricted app developers from directing consumers towards cheaper offers on external websites or alternative app stores.
So Josh, I'm going to come to you first
and is this investigation proof that the EU is finally willing to use its regulatory powers rather than just talk about them, I suppose?
Absolutely.
Yeah.
Yeah.
100%.
This has really empowered Brussels to act swiftly here, which I think is is really necessary in
delivering the promises of consumer protection that came with the the DMA.
I think we do have some precedent before this with big fines for Apple and Meta previously too,
and yeah, this is all in the name of consumer protection.
I think they are coming out guns blazing and using this regulatory power that they have.
Yeah, John, are you surprised by this?
Yeah, I'm surprised and pleased that Google have gone,
I mean, again, it's hard to talk about financial fines when you talk about Google because they have like an infinite amount of money.
Even a, you know, 860 million euro fine, it's like, it's nothing.
I like, you know, the follow-up that with non-compliance, it'll be 5 % of Google's annual turnover.
That feels like a lump sum, something to go at.
But I'll veer into the
less skeptical side of it now and say, it's great to see that the DMA is bearing fruit on this.
It will have an impact on transatlantic relationships, I'm sure there will be threats of tariffs coming, there will be, you know,
all sorts on the back of the DMA and the back of the EU putting this into play.
But it's, yeah, it is good to see that an attempt is being made at least to protect
consumer rights and to do what the DMA was designed to do in the first place.
How might this case reshape how global companies think about compliance, transparency and cross border digital governance?
I think it will make a pretty broad impact.
I think we've seen this previously with the implementation of GDPR and that having effect on companies outside of the governed areas
in terms of how they collect and store and manage consumer personal data.
So I anticipate that this will have very similar knock-on effects
to tech companies is they sort of need to brace for, you know, the most strict regulatory environments, otherwise having to deal with, you know, so many
varying regulations and and privacy laws by by market becomes even even more challenging.
I think the the DMA sort of sets the first standard that
other countries are likely to follow.
They might not take this up as quickly, but I think it's a a strong signal that this is the direction
that Big Tech regulation hopefully is is heading.
And the costs for it as well.
John, do you think tougher EU enforcements could create competitive advantage or a burden?
Ooh, I think that's a tricky one.
Yeah.
No, no, good.
I like it. I've been thrown on the spot.
I think it could be an advantage.
I think as long as it's for the rights of the consumer, I think for a long time that's been the last thing considered.
and you know, the consumers have
gone along with what's been put in front of them, which is totally fair enough.
The rules are very opaque and not easy to understand regarding privacy, regarding what your data is used for, regarding the value exchange of it.
But I think having regulations that show that the best interests of people at heart will always be a decent competitive advantage.
So yeah, I'll fall on that side of the fence today.
I think it is advantageous.
It makes it harder for the companies
to get an easy ride.
And Josh, you're
you're nodding your head there as well in agreement.
I completely agree with everything that John just said.
I think constraint drives innovation.
So these Big Tech companies have nearly infinite resource to figure out how they'll make money and continue to drive record profits in the face of this sort of regulation.
So as long as it's underlined by again protecting consumers,
yeah, I think this is certainly a net positive.
There's no, like you say, they have infinite resource.
There's no excuse not to be able to innovate in these circumstances.
It's not a massive restriction.
It is just doing what is right for your consumers.
And they have the funds to go out there.
They have the people to go out there and do it properly and work around it and innovate around it.
Yeah, innovation.
We like it.
Great.
I'm going to move us on then to our third story for the week.
Paramount Skydance's 110 billion US dollars, equivalent to 86.9 billion pounds,
bid to acquire Warner Bros Discovery has run into a legal obstacle after a US judge temporarily blocked the deal
following a lawsuit from a coalition of 12 state attorney generals.
A US district judge issued a 14 day pause on Monday June 20th after hearing arguments from both sides last week.
The court has set the 3rd of August as a date for a hearing on the state's preliminary injunction motion, although that schedule could actually be pushed back.
Yeah, Josh, do you expect this deal to go through?
I think the odds of this deal going through are certainly getting worse.
Yeah.
And Paramount's made a a real concession here by pushing the the closing date out to 2027.
I think that signals how strong they feel this case is.
So yeah, I think I think the odds are shrinking by the day at this point.
John, what do you think?
I do.
I think I still expect it to go through.
I think the fact it's been cleared by the shareholders and the DOJ have said it's okay is like an indicator that it will eventually close.
But I do, agree with Josh that it's, it is definitely not like a sure thing now.
There is a lot to consider.
Some of the language the judge used around serious questions around antitrust suggests
that they're really taking it seriously in terms of what it could mean for the whole entertainment industry.
And I, yeah, I do personally expect it to still go through.
I just assume these things will.
But I like that this is being so carefully looked at, and that there are, you know, there are reasons for it to be paused,
and that it's being, you know, has that oversight over it. It's a good thing.
Yeah.
If the deal ultimately does proceed, how might a combined Paramount-Warner reshape the landscape of content consolidation and bargaining power with distributors?
Yeah, it certainly gives them uh unprecedented leverage, I think,
and from, you know, a media buying perspective too around this, if we think about how advertisers go into upfronts and how
inventory, premium inventory, is bundled in those upfronts, this just gives
them even more leverage to package up the the good with the bad.
I think ultimately this gives advertisers less choice here really forces their hand.
There's also of course the sort of consumer element to this as well, where yeah the sort of direction for
what gets made, what doesn't get made, probably I guess this really goes beyond consumers.
It goes into anyone working in the industry as well.
There are going to be more hoops to jump through, harder cases to make to get things published that have that like broad reach and,
uh I don't know how I was gonna put a bow on that, but
Yeah, I think, I think this consolidation generally is is going to be harmful.
Yeah.
And John, what do you think a united Paramount-Warner Bros means then for ad buying on TV?
It's a lot of inventory in one place, isn't it?
It's lot of market control, which is why this antitrust, like why this has been paused basically for the antitrust case to be considered.
It's always, it's an analogy I've used a few times, but you know, there used to be six major record labels,
then they, then were five, four, then there were three major record labels.
And it's so much, so much power concentrated in one place and particularly in
streaming is changing TV, everything is fragmenting a little bit, but bring it all back under one house and the power that comes with that.
There's so many questions about that, how that could dominate the ad pricing, how that shifts the market for media buying.
The question around cinema is a big one as well.
That's part of the reason people very anti-Netflix take over because they wouldn't guarantee theatrical release anymore and they want much
shorter windows for cinema stuff compared to TV.
And you wonder how that impacts like the entire entertainment industry really.
It is, yeah, yeah, certainly good that a suitable pause is being taken before this is actually signed off and done.
And all of a sudden, yeah, all this power is concentrated in one place.
I know you kind of touched on consumer concerns, but in terms of market diversity and long-term industry stability,
is there anything else that you think that this challenge raises?
I mean, I would sort of defer to some of the talking points from the unions who are working at entertainment, how like the some of the risks that they've articulated, what
this means for consolidation of power, you know, fewer decision makers really involved in the process and again kind of goes back to what gets made, what doesn't.
Where there is potentially growth in the industry and, you know, ability for more projects to be like greenlit or not.
So yeah I think in the long term this is probably harmful from uh a labor perspective, harmful from an end consumer perspective.
I think yeah, to your point John, the shift to streaming, the impact that that's had on cinema,
there are going to be all these sort of knock on effects of consolidation that probably doesn't work in the best interest
of anyone that sits outside of the Paramount group.
Yeah, yeah.
There's a lot that will change, that's for sure.
And I guess we'll know more in a couple of days time.
So no doubt we'll be diving into this in further detail further or future podcast episodes.
But I think we can leave it there.
And thank you so much, Josh.
Really appreciate you joining us and thank you for sharing your thoughts.
Yeah, thank you so much for having me.
It's been a pleasure.
And John, as always, thank you.
Yeah, I know.
Yeah.
There was a lot.
Excellent.
And thank you everyone for listening and watching.
That's it for The MadTech Podcast this week.
You've been listening to The MadTech Podcast brought to you by ExchangeWire.
We recommend upgrading to the latest Chrome, Firefox, Safari, or Edge.
Please check your internet connection and refresh the page. You might also try disabling any ad blockers.
You can visit our support center if you're having problems.