December 5, 2024 • 34 mins
The UK is in a unique position in the global AI landscape. It is home to important AI development labs and corporate AI adopters, but its regulatory regime is distinct from both the US and the European Union. In this episode, Kevin Werbach sits down with Jessica Leonard, the Chief Strategy and External Affairs Officer at the UK's Competition and Markets Authority (CMA). Jessica discusses the CMA's role in shaping AI policy against the backdrop of a shifting political and economic landscape, and how it balances promoting innovation with competition and consumer protection. She highlights the guiding principles that the CMA has established to ensure a fair and competitive AI ecosystem, and how they are designed to establish trust and fair practices across the industry.
Jessica Lennard took up the role of Chief Strategy & External Affairs Officer at the CMA in August 2023. Jessica is a member of the Senior Executive Team, an advisor to the Board, and has overall responsibility for Strategy, Communications and External Engagement at the CMA. Previously, she was a Senior Director for Global Data and AI Initiatives at VISA. She also served as an Advisory Board Member for the UK Government Centre for Data Ethics and Innovation.
Competition and Markets Authority
CMA AI Strategic Update (April 2024)
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:01):
Jessica, thank you so much for joining me on the road to Accountable AI.
Hi Kevin, how are you?
All right.
It just glitched for a second.
me, let me just do that one more time for some reason it, it just froze for a sec, butprobably just on my end.
But this, this records each one separately.
Let me just do one.
(00:23):
Yeah, for some reason I've got some sort of...
Yeah, it'll be fine.
There's just some sort of...
Jessica, welcome.
Thank you so much for joining me on the road to accountable AI.
Hi Kevin, great to be with you.
First, for those who aren't in the UK or familiar with the system there, what exactly isthe Competition and Markets Authority and what does it have to do with AI policy?
(00:47):
That's a very good question.
Telling me all your listeners don't already know who the CMA is?
Terrible.
So we are the UK Competition and Markets Authority.
We are the primary competition and consumer protection authority for the UK.
So that's obviously a little different to the system you guys have in the US with the DOJand the FTC.
We are two for one.
(01:10):
We are part of the UK government, but we have independent, evidence-led decision-making.
Now, our mandate comes from the UK Parliament.
There's a mandate for us to promote competition within and outside of the UK for thebenefit of consumers.
That's what the legislation says.
(01:30):
And we have a very clear purpose as well, which is our sort of North Star as anorganization.
And that is to promote competitive markets and to tackle unfair behavior.
And we do that for the benefit of people, businesses, and the UK economy.
Now.
I will talk a little bit about the kind of context in which we're doing that in the UK andthen I'll come on to the question around AI.
(01:52):
So we have a new government here in the UK, a new Labour government, and we are having abig, big conversation at the moment about growth in the UK, economic growth.
We have a new national mission, which is a critical priority for the new government todrive that growth that we all want to see across our economy.
Now, for the CMA, growth has been a core pillar of our strategy
(02:15):
for the last couple of years.
our incoming CEO and chair when they came in in early 2023 launched a new strategy for theCMA and productive and sustainable growth across the whole whole UK economy is a core
pillar of that strategy.
So that's really front and center of the way we're thinking about everything that we do atthe moment.
(02:35):
How can we drive that growth, the investment, the innovation, the productivity that weneed in the UK economy right now?
So let me kind of come on to how we're thinking about AI.
So look, we are living, I think it's fair to say, in a state of permanent technologyrevolution these days.
It's very exciting, very disruptive.
And AI is really at the forefront of that.
(02:58):
We are really enthusiastic at the CMA about the potential opportunities that that offers.
We think you have to lean into that, right?
Not just because it has to play a critical part in growth, but also because ultimatelywhat it could lead to is greater prosperity, a transformation for the better in the way
that we live and work, right?
(03:20):
So we're very, very excited about that.
Our role within that is to make sure that that opportunity is fully realized.
Making sure that competition is foundational to that, right?
We've seen time and again, know, history and evidence has shown us that competition canplay a foundational role in that innovation, in that productivity, in driving that
(03:40):
investment, know, unleashing that growth.
So for us, that's really foundational to what we're trying to do in AI.
So what can regulators actually do to promote a domestic AI ecosystem?
I sense some skepticism in your tone, Kevin.
So look, think firstly, there's a big picture piece to all of this, right?
(04:04):
AI is an ecosystem.
It's a big value chain.
And, you know, different regulators will look at different pieces of that, but the bigpicture is very, very important, right?
I think you've got to look upstream at the infrastructure.
You know, what are the kind of, you know, big ticket pieces, the critical inputs we needto...
build models.
And then I think you've got to look downstream at that deployment and release layer andeven below that, know, kind of talent skills, how technology diffuses across an economy,
(04:33):
all of that is incredibly important.
And there are different bits of regulation which might apply, you know, even things likeplanning, immigration, you know, utilities, it's all kind of got a role to play.
So there's a big picture piece.
I think if we're realistic about the UK playing a real role here, becoming a real globalleader,
(04:54):
I think we have to understand, and the CMA is very emphatic about this, that that means arole for companies of all shapes and sizes, right?
From the largest players who've powered a huge amount of innovation and investment in thisspace, through to the scrappy startups and challenges, the specialist providers, and of
course, that huge, fantastic, committed AI ecosystem of researchers and academics,ethicists.
(05:20):
Everyone has got a part to play there.
If I come back to the CMA and our role and our remit, is of course competition.
Again, we think competition can be absolutely foundational to unlocking, unleashing theseopportunities.
And it's not just specific to AI.
History and evidence over a very long time tell us that competition has the power to dothis, to unlock that innovation and all those benefits from market.
(05:50):
I'll give you just four ways in which that happens.
The first is around bringing down barriers to entry and having a level playing field.
If you're a business thinking about starting up or scaling up in a market, you need toknow that you're able to do that in a fair way.
You've got a fair shot at success, at that level playing field.
(06:10):
If you're going to risk your capital, get into something, that level playing field isthere.
And of course, that matters for investors too.
You've got to put your money where you think you're going to get a fair return.
And that level playing field instills confidence in investors, I think, when they'remaking those bets.
Secondly, it's in everybody's interests that the products and services we have are as goodas they can possibly be.
(06:36):
You want companies to succeed on the merits.
You want there to be a continuous cycle of improvement.
And competition is the fire to the feet of companies to make sure that they do that.
Thirdly, and I think this is also very important, for the small minority of bad actors tobe held accountable so that the vast majority of fair dealing businesses have confidence
(06:59):
in that fair shot.
That has to work, and that's also part of competition enforcement.
And then finally, choice, right?
Who doesn't like choice?
You know, consumers and businesses, of course, it's important to have that choice, right?
But again, you know, that diversity of choice, competition can underpin that, not justbecause it helps consumers and businesses, but also there's quite a fundamental point
(07:24):
there around economic resilience.
So a resilient economy, one that is able to withstand shocks and volatility and, know,goodness, that's definitely the
world we live in these days is one in which you have diversity of supply.
We saw that through COVID.
We see it increasingly.
So competition helps to prop up that diversity of supply so that you're not vulnerable tothose single points of failure.
(07:50):
I just want come back to kind of regulation outside of the CMA for a moment and make twopoints.
think the first of those is, as I said at the beginning, there's lots of different areasof regulation that need to be involved here.
This is a general purpose technology.
It's going to impact every part of our society and economy.
So join up is incredibly important.
We'll come on to talk more about how that works in practice later on.
(08:13):
And secondly, regulation should not be the grit in the system, right?
In an innovation led industry like this, that's absolutely fatal, right?
So you need to make sure you're designing it and considering it in ways which understandand take into account the way that innovation led fast moving industries like this work.
So the new government actually here in the UK has just brought in something called theRegulatory Innovation Office.
(08:38):
I think this is really exciting.
It's a specific part of government set up to analyze regulation.
in these sort of innovation led industries and try and work out where it can help andwhere it's hindering and address that.
So I think that's very positive.
Tell us about some of the specific AI issues that you're looking at right now or engagedwith at CMA.
(08:59):
So it's a big field, but I'll talk about just a couple.
So there's a long running area of interest for us around algorithmic harms.
We put out a big piece of research in 2021 on this.
And you'll be familiar with many of these sorts of issues, right?
So, you know, things like AI bias and discrimination, you know, that could be in price orit could be, you know, discrimination against access to certain products and services,
(09:26):
exploitation of vulnerability, you know, and then things like tacit collusion, you know,how can models
kind of play a part in that.
So we continue to be very, very interested in scrutinize those sorts of areas of work.
And we've looked in the last couple of years about things like online choice architecture,dark patterns.
We're looking at how AI can be integrated into those and potentially make some of thoseharms more prevalent.
(09:51):
There's a secondary which we're very excited about at the moment, which is the potentialto
to use AI for competition.
So can we actually, for example, in an area which is really pressing for the UKgovernment, public expenditure, can we use AI here?
So we know that bid rigging is a very real risk in public procurement.
(10:14):
Public procurement is obviously hundreds of billions of pounds worth from the publicpurse.
And we know that if we're going to grow the UK economy, we need investment.
So every penny matters.
Last year alone, the CMA
fine 10 construction companies, 60 million pounds across public and private sectorcontracts for bid rigging.
So we know this is a risk.
(10:36):
We're actually talking to the UK government at the moment about how, with better datasharing and analysis, we could actually use AI to potentially identify that sort of
illegal activity.
So that's obviously very exciting as well.
I will come on then to foundation models.
So for the last two years, this has been a big area of focus for us.
(10:58):
We actually started tracking the market way before two years ago.
We do kind of technology, horizon scanning and insights on a regular basis.
And, you know, it was quite clear to us quite a long time ago that this was an area tokeep an eye on.
And we have done that.
So we do kind of regular tracking and in-depth research into the market.
We started to do a couple of years ago quite extensive stakeholder engagement in thisspace as well.
(11:22):
So we talked to companies of all shapes and sizes all around the world.
We talked to sister authorities and regulators.
We talked to civil society, academics, researchers.
We have some phenomenal folks in the UK, but we also talk all around the world, includingwith plenty of organizations in the US.
And then we've
(11:44):
put out two in-depth reports on this.
The last one was April last year.
So what did we find in all of that research and all of that stakeholder engagement?
Firstly, we found enormous opportunities.
And, as I say, the CMA is just very, very excited about the potential of thesetechnologies.
And that was writ large in the reports that we've published, right?
(12:05):
That transformative potential revolutionized products and services, you know, really likeefficiency and prosperity and just so many things you can see when you delve into the
different use cases here.
Farming, agriculture.
(12:43):
towards positive outcomes.
And that's been very much the frame for us.
How can we use competition and consumer protection to guide markets towards those positiveoutcomes for UK businesses and consumers?
That's our mandate.
And how can we try and mitigate the emergence, the development of some of those risks?
(13:07):
OK, sorry.
Yeah, we might want to cut that question there, and then I'll move on to.
OK.
No, that's fine, we can cut it off at that point,
for just one sec and let me just get my...
Yeah, okay.
Is it worth...
Just a sec, sorry.
(13:29):
Yeah, okay, so I finished that one there.
The question, the 2b one we were gonna come on to about regulation kind of fits there.
Do you wanna...
thought you had taken that with the part you said at the end.
If there's more you want to say, I'm happy to go back to that.
let me pick up just then where I left off of that and say...
(13:50):
Yeah, let me just, yeah, then why don't I ask you the question just to break it up alittle bit, but then, yeah, then can go back to it.
you can say, you know, how did you, so how then are you tackling what you found?
Something like that.
Because I've talked about the risks, so it kind of leads on from there a bit.
So, right, but this is back to the, to the specifically about regulation, so.
(14:10):
Yeah, okay, sure.
Okay.
Okay.
issues that we either we've been looking at.
So how exactly are you tackling those challenges?
Well, I think it's interesting.
Again, if I come back to what not to do.
in some senses, right?
You know, we hear this from companies all the time and I worked in the corporate world forthe vast majority of my career before coming to the CMA.
(14:40):
So I know this to be true, that poorly designed, heavy handed, kind of overbearingregulation really does have the potential to stymie innovation and growth.
And that is not the outcome that the CMA is looking for at all, quite the opposite.
So the approach that we've taken after doing that research
and publishing those reports that I mentioned a moment ago is a principles-based approach.
(15:05):
So we've got six principles that we developed in collaboration with that big stakeholdergroup that I mentioned, consulted on those a lot, and we published the final version of
those with our April report.
So those I can run down for you a little bit.
There's a lot of detail about these available, but I'll give you a quick kind of top line.
So the first one is access.
(15:26):
So ongoing and ready access to the critical inputs that we know are required to build,develop these models.
So things like data, things like compute, things like access to talent.
The second one is diversity.
So we spoke about diversity a moment ago.
You know, the benefits of that diversity in the models that are available, right?
Open source, cloud source.
(15:47):
different capabilities.
We want to see that diversity maintained.
And that leads me into the third principle, which is choice.
You we want the end users and the business customers to have real choice here, not just ofthe different models, but you know, being able to switch across different ecosystems.
We don't want to see any kind of lock in there at all.
(16:08):
The fourth one is fair dealing.
So it's incredibly important in these ecosystems, right?
Where lots of companies of different shapes and sizes have relationships with one another.
And some of those smaller companies really depend on those big ecosystems that there is aprinciple of fair dealing there.
We don't want to be seeing anti-competitive behavior like anti-competitive tying,bundling, self-preferencing, that kind of lock-in.
(16:36):
And then two which are slightly different, so transparency and accountability.
Now, transparency by that we really mean, do end users and business customers have enoughinformation to understand the risks and the limitations of these models?
And accountability, this is really on the firms developing and deploying to make sure thatthey're behaving responsibly, that they're accountable, and that they're upholding that
(17:01):
trust which we know is so incredibly important in this space.
That's kind of the approach that we've taken, not the heavy-handed regulation, but aprinciples-based approach.
And we've been working with that ecosystem on these principles since we published them.
And I have to say the response has been great.
We've been urging companies to embed these into their business practices.
(17:23):
And one of the largest firms, who shall remain nameless, told us recently they thoughtthese principles were the best thing to come out of an antitrust authority last year.
So we'll take that.
Fantastic.
You talked very powerfully earlier about the value of competition in AI, but least to datewhat we've seen, especially for foundation models, is a great deal of concentration really
(17:47):
at every layer of the stack.
So is, is there a viable pathway towards more competition?
I mean, I'm going to say yes, because we would be quite an unambitious competitionauthority if we had a different answer, but it happens to be true, I think.
I mean, again, let's kind of look down the stack a bit.
So in that, you know, those critical inputs that I talked about a moment ago, I think it'svery fair to say that the largest companies are very well positioned here.
(18:13):
You know, they have extremely deep pockets, which is what some of these sort of big ticketitems require.
And I think it's also important to recognize the contribution they've made here, actually,you know, to R &D, to investment, to talent.
You know, they've been absolute engines of that stuff.
So, you know, that's, think, very important to recognize.
(18:35):
Does that mean there is not also a role for the smaller players?
No, I think very, very much there is.
Right.
And it's also the CMA's job to make sure that that materializes.
And I think that's I think it's very, you know, when you look at the last couple of years,
right, the sort of curve jumping innovation we've seen.
Some of the most innovative and capable models have actually come from the smallerplayers, models that are now becoming quite quickly household names, Claude, Perplexity,
(19:02):
OpenAI, even my mom knows now what Perplexity is.
So I make sure she's quite tech savvy.
I think what that kind of shows you is that there really is kind of room for everybodyhere as long as we do keep in mind and really try to take seriously the principles that
I just mentioned.
Now, of course, competition is what underpins the sustainability of what I've justdescribed, right?
(19:28):
That kind of diversity in the ecosystem and that healthy tension that keeps companiescompeting with one another, right?
The large ones with one another, and you do see that increasingly in this space too, aswell as the smaller ones.
So our job is very much to try and safeguard and enhance that, right?
(19:49):
I think there is also...
a layer down in the deployment and release layer that I'll come to talk a little bit moreabout competition there when I talk about the new digital markets competition and
consumers act in the UK.
But I do also just want to sort of come back a bit to the question of how is thispossible?
How are we going to maintain these things?
(20:10):
And a piece I talked about a moment ago in terms of regulation.
if we are able to keep moving in the right direction, to keep this diversity, to keep thiskind of competitive rivalry going.
What that kind of helps you avoid is needing to unpick some of this stuff later on downthe line when harms have potentially become really entrenched.
(20:33):
So early action, working together to do that can actually help avoid some of the sort ofmore disruptive type of interventions, right?
The sort of regulation that we know companies don't want layered on top of yet moreregulation.
So we think there's a very positive way to move forward as an ecosystem.
(20:54):
on all of this and hopefully prevent those problems from becoming entrenched and requiringmore costly, more disruptive intervention later on.
How do you develop the kinds of internal capacity and expertise though that you need toaddress such a technologically sophisticated area?
It's a very, very good question.
(21:14):
And again, you know, coming from the private sector, I've been astonished at how well theCMA does this.
You know, they really did get ahead of the game in doing this.
So in 2019, we set up our data technology and insights unit.
And this is now an 80 strong multidisciplinary cross CMA function.
(21:36):
It's absolutely fantastic.
We've got data scientists, data engineers, e-discovery, forensics, behavioral scientists,and of
know, increasingly lawyers and economists and policy folk as well, kind of, you know, allreally embedded in this.
So it's a fantastic capability.
I'll say 80 again, just in case any of your listeners miss her and thought I said 800 or8000.
(22:00):
But you know, nonetheless, 80, I think puts us really at the forefront of these sort ofcapabilities for comparable authorities around the world.
And it is incredibly important to have that right, you know, these are complicated, fastmoving markets.
So the onus is
on us to make sure we really know what we're doing in this space.
Now, it's not enough for us just to be doing that internally.
(22:21):
The second piece that I would really emphasize is you don't learn unless you listen withsomething as technical and complicated as this.
So that stakeholder engagement I've been talking about is really important.
So Sarah Cardell, when she came in as CEO, she's made a very, very conscious push for theCMA to be really listening, engaging more than we have ever engaged with companies.
(22:43):
We have a much more open door policy now for companies of all shapes and sizes to come andtalk to us, explain their business models, understand us and our processes.
And I would really urge companies to do that, actually.
That's something we're very, committed to.
So that listening and engaging.
Also, we've got kind of, know,
experience here which serves us in very good stead.
(23:05):
So the work we've done around mobile ecosystems, around digital advertising, around cloud,online choice architecture, I talked about dark patterns.
Some of this is market studies and some of it is really in-depth investigation, some isresearch, but there's a really rich kind of body of expertise and experience that has been
built up over quite a significant number of years now, which we bring to bear when we lookat things like AI.
(23:34):
Are there any open investigations or enforcement proceedings in the US, for example, theFederal Trade Commission has an investigation ongoing of open AI?
Is there anything similar happening in the UK?
So we've thought very carefully across what we call our toolkit, the sort of suite of ourpowers.
And it does work a little differently here than it does in the US.
(23:55):
So we've kind of thought across the suite of what we have available to us.
So I'll give you kind of a bit of a rundown of what we've been doing and thinking about.
So firstly, we've been doing the research and the reports that I mentioned earlier andthat kind of ecosystem engagement.
We're looking at AI through the lens of our markets work.
So we have this quite cool tool here in the UK called a market investigation.
(24:16):
This is a really in-depth look at a market led by an internal, sorry, an independent groupof experts.
This is really forensic evidence led process.
So we have our cloud market investigation open at the moment.
And we have said that we will as part of that be considering the impact of foundationmodels on competition in cloud.
(24:40):
services.
So that's kind of one way that we can look at these things.
And yes, we have a mergers tool.
And, and we've been kind of thinking about AI in the context of that as well.
So just to sort of frame that a little bit and bring us back to what we were talking aboutearlier in terms of the CMA role.
So again, you know, when we're looking at all of this in the round, we're thinking aboutwhat how can we as a competition authority unlock that opportunity of the AI of AI for the
(25:08):
UK, right?
How can how can competition under
pin growth in these markets.
And we're also thinking about that duty I talked about, which is given to us byParliament, to promote competition within and outside of the UK for the benefit of
consumers.
So with both of those in mind, we have been taking a look at a handful of partnershipsthat have been taking place across the AI ecosystem between some of the largest players
(25:35):
and some of those smaller players that I spoke about earlier.
And you know, some of those deals are a little complex and a little opaque.
It's not immediately obvious with some of them what's going on.
we look at that with two questions in mind, which are based on how merger law works herein the UK.
(25:55):
So the first of those is, do we have jurisdiction to look at the deal?
Does it constitute a deal under UK merger law?
And secondly, does it really raise competition concerns?
Do we think there's likelihood of a substantial lessening of competition in the UK?
So that's kind of the frame through which we've been looking at these deals.
Now, when I say looking at them, what that tends to mean is we have a phase one process,which is really a sort of information and evidence gathering process that gives the
(26:25):
parties themselves the opportunity to submit documents.
And they do a lot.
But also third parties, know, anyone can kind of submit evidence to this process.
It will all be looked at in the round by experts within the CMA.
So we've been running through this handful of cases in a, I think, a very timely andefficient way, looking at that evidence.
(26:49):
And as we go, we've been really striving to give greater clarity.
You know, we've been ourselves kind of hoping to get greater clarity around the sorts ofdeals that do and don't meet the two
criteria that I spoke about earlier.
And also, we've been publishing very fulsome decisions.
Not every authority around the world publishes a full decision if it's a clearance, right?
(27:12):
We do do that.
So we've been really trying to unpack our thinking in there because we know how importantit is for businesses, right, to have that clarity, especially investors.
So we've been really trying to do that.
And, and we've been actually publishing supplementary commentary on top just to just totry and
just to try and be as transparent as we possibly can.
(27:35):
And that's really the way that the phase one process is designed to work here in the UK.
It's a little unfamiliar perhaps for folks across the pond.
The other thing I would say just to put all of that into context is there are around50,000 &A deals announced every year.
Here in the UK, we took a look last year at 913 of those.
(27:58):
54 we investigated.
And one was blocked.
Just one.
So I think, you know, it's very fair to say that the way that we treat this is consciouslyproportionate.
That's the way the legislation is designed.
There's a very, very high bar.
And that's certainly our attitude when we look at these things.
(28:20):
What do you see as the major open issues or open questions right now for AI?
So, you we've talked about the competition piece quite a lot.
So let me sort of move away from that and talk about another couple of issues that Ithink, you know, should be on everybody's radar.
(28:41):
So the first one of those is the other half of our remit, which is consumer protection.
You know, this is this is very much in everyone's minds.
I'm sure that it is with yours, given the name of your podcast.
As I say, algorithmic harms have been, you know, something on our minds for some time.
And, you know, the report that we put out in April also looked quite in depth at thepropensity of AI to exacerbate those harms.
(29:05):
Now, you know, we also know that AI has incredible potential to do fantastic
things, right, for consumers.
But I think we want to be thoughtful about those two principles I mentioned of the CMAs,transparency and accountability, particularly in this context.
I think it's incredibly important.
(29:26):
Secondly, I think AI safety is a big one, right, that, you know, we should talk moreabout.
This is one which is kind of, you know,
catapulted up the political agenda for good reasons in the last couple of years.
And I think it's extremely important that we maintain the global momentum that we have nowbehind this.
(29:51):
And then the other piece I would sort of add on a little bit is what I talked about at thebeginning in terms of all of the different areas of regulation that need to work together
here.
So it's quite easy, I think, for all of us in AI to lose sight of the bigger picture
But you hear a lot at the moment around energy, for example, and the importance of thatwithin the AI ecosystem.
(30:13):
We know that it's incredibly important that infrastructure and planning work in the rightway in order to facilitate the growth of these industries.
So there's a big suite of policy issues there.
Absolutely.
I'm going to skip the question about coordinating with other agencies just for time, ifthat's okay.
Yeah, so we're just going.
So I'll do the DMCCA one next.
(30:42):
You mentioned earlier the DMCCA, the UK Digital Markets Competition and Consumers Act,which is different from the Digital Markets Act, the DMA in the European Union.
So could you say a little bit more about how that legislation impacts your authority andability to address these kinds of issues?
(31:04):
It is different, it is different.
And I think it's worth noting that, you know, whilst
authorities all over the world are taking different approaches to this.
Many of the principles are the same, right?
I think there's been a broad acknowledgement that, you know, for an innovation-led,fast-paced markets like digital, the sort of regulation that we had for quite a long time
(31:27):
just wasn't going to cut it, right?
So different places around the world have been developing their own kind of responses tothat.
And the digital markets competition and consumer regime in the UK is ours.
We like to
think that it is a very, very well designed and unique piece of regulation.
There's a few features which make it so, I'll talk to you about those again.
(31:51):
coming back to the principle I described earlier about how competition can unleash andunlock growth and investment, this is also the basis for this piece of regulation.
The idea here is to maximize the opportunities to grow a UK tech sector.
All right.
And that's incredibly exciting.
(32:12):
So that should mean that opportunities for the largest companies remain, right?
They play an incredibly important role in these digital markets.
And, you know, that should remain, but it means that the CMA will also have an ability toprotect and enhance opportunities for the smaller players, right?
For those scrappy startups, for the challenges, right?
(32:33):
To come in and have that fair shot that I talked about earlier.
And it should also mean fantastic opportunities for consumers and business customers toget great choices and fair deals.
So that's kind of the framing context for this.
We're hoping it's going to be genuinely a new era in digital growth for the UK economy.
(32:53):
So how is it going to do that exactly?
So it is supposed to apply to only a very small number of the very largest firms.
So those firms will have what's called strategic market status.
So this is a bit legalistic, but effectively what it means is they will have a substantialand entrenched position in one or more digital markets in the UK.
(33:20):
Now we haven't made any provisional decisions yet about where we're to focus thoseinvestigations.
But what I can talk about a bit is the principles and the process underpinning the waythat this regulation will work and why we think it's so special and why we think it should
be extremely attractive actually to
to businesses and investors in the way that it operates.
(33:40):
So first of all, it's designed to be very tailored and bespoke.
It's not a big sweeping, one size fits all piece of regulation.
We think that's really, really important for the reasons that I've described earlier.
Instead, what it allows us to do is really target specific homes with tailored solutions,working iteratively with the companies in a participative conversation that involves the
(34:05):
whole ecosystem, a wide range of stakeholders, so that the interventions we're makingreally are targeted to exactly that problem.
And it allows us to be quite iterative and bespoke in the way that we're doing it and veryproportionate as well.
This is a truly different approach to what I think we see in other pieces of equivalentregulation.
(34:27):
And certainly, this is something that we think is going to be very attractive, actually,to businesses.
And certainly, we hope to the consumers that will also benefit from all of this.
Now, AI is a consideration within that.
We've kind of talked about that a little bit.
How that might work in practice, for example, is where AI is part of one of the criticalinputs that we might look at in one of those digital activities.
(34:57):
cloud, for example.
Equally, in that deployment layer that I said I'd come back to when you're looking atthings like mobile ecosystems, search, productivity software, these are all potential
avenues for foundation models to be deployed and released across the economy.
That is also something that we can look at through the lens of this regulation.
(35:21):
Would you say that you've learned anything even before the DMCCA?
Yeah, that's okay.
That's okay.
While you're doing that, I'm just going to check something very, very quickly.
Yeah.
then we're almost done.
Just two more questions.
(35:42):
Shoot, you know, I misspoke and I knew I misspoke when I said it.
Yeah, I tell you what I'm going to do super quick.
I'm going to just say at the end of what I was talking about mergers, I said, I'm going tostep back for some context here.
There are 50,000 deals, right?
I'm just going to do that one little bit again if we can.
(36:03):
Thank you so much.
Damn lies and statistics.
Right.
Okay, let me just reverse for a second.
Yes.
Okay.
I think it's also worth just sort of noting for context here that around 50,000 &A dealswere announced last year.
(36:26):
And the CMA looked at 913.
Now 95 % of those didn't even proceed to an investigation, let alone an intervention ofsome kind.
And in fact, just one was prohibited.
So it's quite useful to keep those sort of figures in mind.
(36:46):
Super, thank you.
All right.
And how do you want to, where do you want to go next?
So I was just gonna ask you, what have you learned from, even pre-DMCCA, what have youlearned from the experience trying to exercise oversight over the big platforms?
God.
If you don't, I mean if that doesn't sound right, we could, you know, we've got enoughtime.
be sort of slightly encroaching on Will's territory there.
(37:10):
That's fine.
No, don't worry about it.
We've got enough content.
So why don't I just go on to the last one, which is the global piece.
That's totally fine.
Okay.
Because yeah, no, we've got, mean, there's, this is more than enough time even without anythings.
Great, okay.
So last question, the UK hosted a major high level AI safety summit in late 2023 thatresulted in an international declaration signed by I think 28 different countries around
(37:42):
AI policy.
To what extent do you think there is potential going forward for truly global governanceof this phenomenon given how global the technology is?
Very good question.
makes me feel a little tired when you ask it like that.
In my previous job, I had a big glass wall.
(38:04):
And on that glass wall, every time I found a new AI ethics or governance framework, Iwould just note it down and I ran out of wall really quite quickly.
So, you when you think about kind of global harmonization, yes, it seems it seems a longway off and to an extent that there are good reasons for that, right?
know, nations are sovereign in terms of the, you know, the legal frameworks that they haveand the decisions that they take, you know.
(38:30):
different countries are at different stages of technological development, have differentsocial norms.
So, you know, that I think is understandable and healthy.
I think when you when you think about an issue like this, and you look at the sort ofexistential risks that were discussed at Bletchley and in those sorts of conversations, it
(38:50):
is eminently important that we do have that sort of global consensus and globalconversation.
I think Bletchley was incredibly important in that.
it's fantastic that that
is continuing around the world.
from a competition perspective, we're a really active part of our international communitytalking about these things.
know, groups like the G7 just issued a communique last month on a commitment in this area.
(39:18):
Equally, we put out a joint statement over the summer with the FTC, the DOJ and theEuropean Commission around upholding competition and consumer protection in the AI space.
So we can all be we can all be doing that and, you know, continuing to to kind of
share best practice to support one another in terms of understanding and sharing the kindof capabilities and expertise that you need in this space.
(39:43):
These are borderless markets, right?
And we live in a very globally interconnected world.
So those relationships are strong already.
They've been getting stronger over the last several decades.
And I think they will continue to strengthen around an issue like AI.
I mean, it's interesting, we were talking earlier about the different approaches even todigital regulation being taken around the world.
(40:08):
In the EU now they have a big AI act.
In the US, I know it's kind of a bit of a fragmented picture.
Last I heard California's got 32 AI laws of its own.
And here in the UK, I say, we're, CMA is gonna take this principles-based approach.
Government's approach is kind of look at this through the lens of sectorial regulationprimarily, and then it has
(40:30):
the AI safety work as a kind of different strand and it will be bringing forward an AIsafety bill within this parliament.
So all of these different approaches, I hope, add up to some alignment on the need forreal action in this space.
But true harmonization, think we may need to come back in a few years and do anotherpodcast on that.
(40:53):
Well, that's all right.
Covered a lot and this is a very fast moving area.
Jessica, really appreciate your taking the time to speak with me.
Not at all.
Thanks so much, Kevin.
Jessica, thank you so much for joining me on the road to Accountable AI.
Hi Kevin, how are you?
All right.
It just glitched for a second.
me, let me just do that one more time for some reason it, it just froze for a sec, butprobably just on my end.
But this, this records each one separately.
Let me just do one.
(00:23):
Yeah, for some reason I've got some sort of...
Yeah, it'll be fine.
There's just some sort of...
Jessica, welcome.
Thank you so much for joining me on the road to accountable AI.
Hi Kevin, great to be with you.
First, for those who aren't in the UK or familiar with the system there, what exactly isthe Competition and Markets Authority and what does it have to do with AI policy?
(00:47):
That's a very good question.
Telling me all your listeners don't already know who the CMA is?
Terrible.
So we are the UK Competition and Markets Authority.
We are the primary competition and consumer protection authority for the UK.
So that's obviously a little different to the system you guys have in the US with the DOJand the FTC.
We are two for one.
(01:10):
We are part of the UK government, but we have independent, evidence-led decision-making.
Now, our mandate comes from the UK Parliament.
There's a mandate for us to promote competition within and outside of the UK for thebenefit of consumers.
That's what the legislation says.
(01:30):
And we have a very clear purpose as well, which is our sort of North Star as anorganization.
And that is to promote competitive markets and to tackle unfair behavior.
And we do that for the benefit of people, businesses, and the UK economy.
Now.
I will talk a little bit about the kind of context in which we're doing that in the UK andthen I'll come on to the question around AI.
(01:52):
So we have a new government here in the UK, a new Labour government, and we are having abig, big conversation at the moment about growth in the UK, economic growth.
We have a new national mission, which is a critical priority for the new government todrive that growth that we all want to see across our economy.
Now, for the CMA, growth has been a core pillar of our strategy
(02:15):
for the last couple of years.
our incoming CEO and chair when they came in in early 2023 launched a new strategy for theCMA and productive and sustainable growth across the whole whole UK economy is a core
pillar of that strategy.
So that's really front and center of the way we're thinking about everything that we do atthe moment.
(02:35):
How can we drive that growth, the investment, the innovation, the productivity that weneed in the UK economy right now?
So let me kind of come on to how we're thinking about AI.
So look, we are living, I think it's fair to say, in a state of permanent technologyrevolution these days.
It's very exciting, very disruptive.
And AI is really at the forefront of that.
(02:58):
We are really enthusiastic at the CMA about the potential opportunities that that offers.
We think you have to lean into that, right?
Not just because it has to play a critical part in growth, but also because ultimatelywhat it could lead to is greater prosperity, a transformation for the better in the way
that we live and work, right?
(03:20):
So we're very, very excited about that.
Our role within that is to make sure that that opportunity is fully realized.
Making sure that competition is foundational to that, right?
We've seen time and again, know, history and evidence has shown us that competition canplay a foundational role in that innovation, in that productivity, in driving that
(03:40):
investment, know, unleashing that growth.
So for us, that's really foundational to what we're trying to do in AI.
So what can regulators actually do to promote a domestic AI ecosystem?
I sense some skepticism in your tone, Kevin.
So look, think firstly, there's a big picture piece to all of this, right?
(04:04):
AI is an ecosystem.
It's a big value chain.
And, you know, different regulators will look at different pieces of that, but the bigpicture is very, very important, right?
I think you've got to look upstream at the infrastructure.
You know, what are the kind of, you know, big ticket pieces, the critical inputs we needto...
build models.
And then I think you've got to look downstream at that deployment and release layer andeven below that, know, kind of talent skills, how technology diffuses across an economy,
(04:33):
all of that is incredibly important.
And there are different bits of regulation which might apply, you know, even things likeplanning, immigration, you know, utilities, it's all kind of got a role to play.
So there's a big picture piece.
I think if we're realistic about the UK playing a real role here, becoming a real globalleader,
(04:54):
I think we have to understand, and the CMA is very emphatic about this, that that means arole for companies of all shapes and sizes, right?
From the largest players who've powered a huge amount of innovation and investment in thisspace, through to the scrappy startups and challenges, the specialist providers, and of
course, that huge, fantastic, committed AI ecosystem of researchers and academics,ethicists.
(05:20):
Everyone has got a part to play there.
If I come back to the CMA and our role and our remit, is of course competition.
Again, we think competition can be absolutely foundational to unlocking, unleashing theseopportunities.
And it's not just specific to AI.
History and evidence over a very long time tell us that competition has the power to dothis, to unlock that innovation and all those benefits from market.
(05:50):
I'll give you just four ways in which that happens.
The first is around bringing down barriers to entry and having a level playing field.
If you're a business thinking about starting up or scaling up in a market, you need toknow that you're able to do that in a fair way.
You've got a fair shot at success, at that level playing field.
(06:10):
If you're going to risk your capital, get into something, that level playing field isthere.
And of course, that matters for investors too.
You've got to put your money where you think you're going to get a fair return.
And that level playing field instills confidence in investors, I think, when they'remaking those bets.
Secondly, it's in everybody's interests that the products and services we have are as goodas they can possibly be.
(06:36):
You want companies to succeed on the merits.
You want there to be a continuous cycle of improvement.
And competition is the fire to the feet of companies to make sure that they do that.
Thirdly, and I think this is also very important, for the small minority of bad actors tobe held accountable so that the vast majority of fair dealing businesses have confidence
(06:59):
in that fair shot.
That has to work, and that's also part of competition enforcement.
And then finally, choice, right?
Who doesn't like choice?
You know, consumers and businesses, of course, it's important to have that choice, right?
But again, you know, that diversity of choice, competition can underpin that, not justbecause it helps consumers and businesses, but also there's quite a fundamental point
(07:24):
there around economic resilience.
So a resilient economy, one that is able to withstand shocks and volatility and, know,goodness, that's definitely the
world we live in these days is one in which you have diversity of supply.
We saw that through COVID.
We see it increasingly.
So competition helps to prop up that diversity of supply so that you're not vulnerable tothose single points of failure.
(07:50):
I just want come back to kind of regulation outside of the CMA for a moment and make twopoints.
think the first of those is, as I said at the beginning, there's lots of different areasof regulation that need to be involved here.
This is a general purpose technology.
It's going to impact every part of our society and economy.
So join up is incredibly important.
We'll come on to talk more about how that works in practice later on.
(08:13):
And secondly, regulation should not be the grit in the system, right?
In an innovation led industry like this, that's absolutely fatal, right?
So you need to make sure you're designing it and considering it in ways which understandand take into account the way that innovation led fast moving industries like this work.
So the new government actually here in the UK has just brought in something called theRegulatory Innovation Office.
(08:38):
I think this is really exciting.
It's a specific part of government set up to analyze regulation.
in these sort of innovation led industries and try and work out where it can help andwhere it's hindering and address that.
So I think that's very positive.
Tell us about some of the specific AI issues that you're looking at right now or engagedwith at CMA.
(08:59):
So it's a big field, but I'll talk about just a couple.
So there's a long running area of interest for us around algorithmic harms.
We put out a big piece of research in 2021 on this.
And you'll be familiar with many of these sorts of issues, right?
So, you know, things like AI bias and discrimination, you know, that could be in price orit could be, you know, discrimination against access to certain products and services,
(09:26):
exploitation of vulnerability, you know, and then things like tacit collusion, you know,how can models
kind of play a part in that.
So we continue to be very, very interested in scrutinize those sorts of areas of work.
And we've looked in the last couple of years about things like online choice architecture,dark patterns.
We're looking at how AI can be integrated into those and potentially make some of thoseharms more prevalent.
(09:51):
There's a secondary which we're very excited about at the moment, which is the potentialto
to use AI for competition.
So can we actually, for example, in an area which is really pressing for the UKgovernment, public expenditure, can we use AI here?
So we know that bid rigging is a very real risk in public procurement.
(10:14):
Public procurement is obviously hundreds of billions of pounds worth from the publicpurse.
And we know that if we're going to grow the UK economy, we need investment.
So every penny matters.
Last year alone, the CMA
fine 10 construction companies, 60 million pounds across public and private sectorcontracts for bid rigging.
So we know this is a risk.
(10:36):
We're actually talking to the UK government at the moment about how, with better datasharing and analysis, we could actually use AI to potentially identify that sort of
illegal activity.
So that's obviously very exciting as well.
I will come on then to foundation models.
So for the last two years, this has been a big area of focus for us.
(10:58):
We actually started tracking the market way before two years ago.
We do kind of technology, horizon scanning and insights on a regular basis.
And, you know, it was quite clear to us quite a long time ago that this was an area tokeep an eye on.
And we have done that.
So we do kind of regular tracking and in-depth research into the market.
We started to do a couple of years ago quite extensive stakeholder engagement in thisspace as well.
(11:22):
So we talked to companies of all shapes and sizes all around the world.
We talked to sister authorities and regulators.
We talked to civil society, academics, researchers.
We have some phenomenal folks in the UK, but we also talk all around the world, includingwith plenty of organizations in the US.
And then we've
(11:44):
put out two in-depth reports on this.
The last one was April last year.
So what did we find in all of that research and all of that stakeholder engagement?
Firstly, we found enormous opportunities.
And, as I say, the CMA is just very, very excited about the potential of thesetechnologies.
And that was writ large in the reports that we've published, right?
(12:05):
That transformative potential revolutionized products and services, you know, really likeefficiency and prosperity and just so many things you can see when you delve into the
different use cases here.
Farming, agriculture.
(12:43):
towards positive outcomes.
And that's been very much the frame for us.
How can we use competition and consumer protection to guide markets towards those positiveoutcomes for UK businesses and consumers?
That's our mandate.
And how can we try and mitigate the emergence, the development of some of those risks?
(13:07):
OK, sorry.
Yeah, we might want to cut that question there, and then I'll move on to.
OK.
No, that's fine, we can cut it off at that point,
for just one sec and let me just get my...
Yeah, okay.
Is it worth...
Just a sec, sorry.
(13:29):
Yeah, okay, so I finished that one there.
The question, the 2b one we were gonna come on to about regulation kind of fits there.
Do you wanna...
thought you had taken that with the part you said at the end.
If there's more you want to say, I'm happy to go back to that.
let me pick up just then where I left off of that and say...
(13:50):
Yeah, let me just, yeah, then why don't I ask you the question just to break it up alittle bit, but then, yeah, then can go back to it.
you can say, you know, how did you, so how then are you tackling what you found?
Something like that.
Because I've talked about the risks, so it kind of leads on from there a bit.
So, right, but this is back to the, to the specifically about regulation, so.
(14:10):
Yeah, okay, sure.
Okay.
Okay.
issues that we either we've been looking at.
So how exactly are you tackling those challenges?
Well, I think it's interesting.
Again, if I come back to what not to do.
in some senses, right?
You know, we hear this from companies all the time and I worked in the corporate world forthe vast majority of my career before coming to the CMA.
(14:40):
So I know this to be true, that poorly designed, heavy handed, kind of overbearingregulation really does have the potential to stymie innovation and growth.
And that is not the outcome that the CMA is looking for at all, quite the opposite.
So the approach that we've taken after doing that research
and publishing those reports that I mentioned a moment ago is a principles-based approach.
(15:05):
So we've got six principles that we developed in collaboration with that big stakeholdergroup that I mentioned, consulted on those a lot, and we published the final version of
those with our April report.
So those I can run down for you a little bit.
There's a lot of detail about these available, but I'll give you a quick kind of top line.
So the first one is access.
(15:26):
So ongoing and ready access to the critical inputs that we know are required to build,develop these models.
So things like data, things like compute, things like access to talent.
The second one is diversity.
So we spoke about diversity a moment ago.
You know, the benefits of that diversity in the models that are available, right?
Open source, cloud source.
(15:47):
different capabilities.
We want to see that diversity maintained.
And that leads me into the third principle, which is choice.
You we want the end users and the business customers to have real choice here, not just ofthe different models, but you know, being able to switch across different ecosystems.
We don't want to see any kind of lock in there at all.
(16:08):
The fourth one is fair dealing.
So it's incredibly important in these ecosystems, right?
Where lots of companies of different shapes and sizes have relationships with one another.
And some of those smaller companies really depend on those big ecosystems that there is aprinciple of fair dealing there.
We don't want to be seeing anti-competitive behavior like anti-competitive tying,bundling, self-preferencing, that kind of lock-in.
(16:36):
And then two which are slightly different, so transparency and accountability.
Now, transparency by that we really mean, do end users and business customers have enoughinformation to understand the risks and the limitations of these models?
And accountability, this is really on the firms developing and deploying to make sure thatthey're behaving responsibly, that they're accountable, and that they're upholding that
(17:01):
trust which we know is so incredibly important in this space.
That's kind of the approach that we've taken, not the heavy-handed regulation, but aprinciples-based approach.
And we've been working with that ecosystem on these principles since we published them.
And I have to say the response has been great.
We've been urging companies to embed these into their business practices.
(17:23):
And one of the largest firms, who shall remain nameless, told us recently they thoughtthese principles were the best thing to come out of an antitrust authority last year.
So we'll take that.
Fantastic.
You talked very powerfully earlier about the value of competition in AI, but least to datewhat we've seen, especially for foundation models, is a great deal of concentration really
(17:47):
at every layer of the stack.
So is, is there a viable pathway towards more competition?
I mean, I'm going to say yes, because we would be quite an unambitious competitionauthority if we had a different answer, but it happens to be true, I think.
I mean, again, let's kind of look down the stack a bit.
So in that, you know, those critical inputs that I talked about a moment ago, I think it'svery fair to say that the largest companies are very well positioned here.
(18:13):
You know, they have extremely deep pockets, which is what some of these sort of big ticketitems require.
And I think it's also important to recognize the contribution they've made here, actually,you know, to R &D, to investment, to talent.
You know, they've been absolute engines of that stuff.
So, you know, that's, think, very important to recognize.
(18:35):
Does that mean there is not also a role for the smaller players?
No, I think very, very much there is.
Right.
And it's also the CMA's job to make sure that that materializes.
And I think that's I think it's very, you know, when you look at the last couple of years,
right, the sort of curve jumping innovation we've seen.
Some of the most innovative and capable models have actually come from the smallerplayers, models that are now becoming quite quickly household names, Claude, Perplexity,
(19:02):
OpenAI, even my mom knows now what Perplexity is.
So I make sure she's quite tech savvy.
I think what that kind of shows you is that there really is kind of room for everybodyhere as long as we do keep in mind and really try to take seriously the principles that
I just mentioned.
Now, of course, competition is what underpins the sustainability of what I've justdescribed, right?
(19:28):
That kind of diversity in the ecosystem and that healthy tension that keeps companiescompeting with one another, right?
The large ones with one another, and you do see that increasingly in this space too, aswell as the smaller ones.
So our job is very much to try and safeguard and enhance that, right?
(19:49):
I think there is also...
a layer down in the deployment and release layer that I'll come to talk a little bit moreabout competition there when I talk about the new digital markets competition and
consumers act in the UK.
But I do also just want to sort of come back a bit to the question of how is thispossible?
How are we going to maintain these things?
(20:10):
And a piece I talked about a moment ago in terms of regulation.
if we are able to keep moving in the right direction, to keep this diversity, to keep thiskind of competitive rivalry going.
What that kind of helps you avoid is needing to unpick some of this stuff later on downthe line when harms have potentially become really entrenched.
(20:33):
So early action, working together to do that can actually help avoid some of the sort ofmore disruptive type of interventions, right?
The sort of regulation that we know companies don't want layered on top of yet moreregulation.
So we think there's a very positive way to move forward as an ecosystem.
(20:54):
on all of this and hopefully prevent those problems from becoming entrenched and requiringmore costly, more disruptive intervention later on.
How do you develop the kinds of internal capacity and expertise though that you need toaddress such a technologically sophisticated area?
It's a very, very good question.
(21:14):
And again, you know, coming from the private sector, I've been astonished at how well theCMA does this.
You know, they really did get ahead of the game in doing this.
So in 2019, we set up our data technology and insights unit.
And this is now an 80 strong multidisciplinary cross CMA function.
(21:36):
It's absolutely fantastic.
We've got data scientists, data engineers, e-discovery, forensics, behavioral scientists,and of
know, increasingly lawyers and economists and policy folk as well, kind of, you know, allreally embedded in this.
So it's a fantastic capability.
I'll say 80 again, just in case any of your listeners miss her and thought I said 800 or8000.
(22:00):
But you know, nonetheless, 80, I think puts us really at the forefront of these sort ofcapabilities for comparable authorities around the world.
And it is incredibly important to have that right, you know, these are complicated, fastmoving markets.
So the onus is
on us to make sure we really know what we're doing in this space.
Now, it's not enough for us just to be doing that internally.
(22:21):
The second piece that I would really emphasize is you don't learn unless you listen withsomething as technical and complicated as this.
So that stakeholder engagement I've been talking about is really important.
So Sarah Cardell, when she came in as CEO, she's made a very, very conscious push for theCMA to be really listening, engaging more than we have ever engaged with companies.
(22:43):
We have a much more open door policy now for companies of all shapes and sizes to come andtalk to us, explain their business models, understand us and our processes.
And I would really urge companies to do that, actually.
That's something we're very, committed to.
So that listening and engaging.
Also, we've got kind of, know,
experience here which serves us in very good stead.
(23:05):
So the work we've done around mobile ecosystems, around digital advertising, around cloud,online choice architecture, I talked about dark patterns.
Some of this is market studies and some of it is really in-depth investigation, some isresearch, but there's a really rich kind of body of expertise and experience that has been
built up over quite a significant number of years now, which we bring to bear when we lookat things like AI.
(23:34):
Are there any open investigations or enforcement proceedings in the US, for example, theFederal Trade Commission has an investigation ongoing of open AI?
Is there anything similar happening in the UK?
So we've thought very carefully across what we call our toolkit, the sort of suite of ourpowers.
And it does work a little differently here than it does in the US.
(23:55):
So we've kind of thought across the suite of what we have available to us.
So I'll give you kind of a bit of a rundown of what we've been doing and thinking about.
So firstly, we've been doing the research and the reports that I mentioned earlier andthat kind of ecosystem engagement.
We're looking at AI through the lens of our markets work.
So we have this quite cool tool here in the UK called a market investigation.
(24:16):
This is a really in-depth look at a market led by an internal, sorry, an independent groupof experts.
This is really forensic evidence led process.
So we have our cloud market investigation open at the moment.
And we have said that we will as part of that be considering the impact of foundationmodels on competition in cloud.
(24:40):
services.
So that's kind of one way that we can look at these things.
And yes, we have a mergers tool.
And, and we've been kind of thinking about AI in the context of that as well.
So just to sort of frame that a little bit and bring us back to what we were talking aboutearlier in terms of the CMA role.
So again, you know, when we're looking at all of this in the round, we're thinking aboutwhat how can we as a competition authority unlock that opportunity of the AI of AI for the
(25:08):
UK, right?
How can how can competition under
pin growth in these markets.
And we're also thinking about that duty I talked about, which is given to us byParliament, to promote competition within and outside of the UK for the benefit of
consumers.
So with both of those in mind, we have been taking a look at a handful of partnershipsthat have been taking place across the AI ecosystem between some of the largest players
(25:35):
and some of those smaller players that I spoke about earlier.
And you know, some of those deals are a little complex and a little opaque.
It's not immediately obvious with some of them what's going on.
we look at that with two questions in mind, which are based on how merger law works herein the UK.
(25:55):
So the first of those is, do we have jurisdiction to look at the deal?
Does it constitute a deal under UK merger law?
And secondly, does it really raise competition concerns?
Do we think there's likelihood of a substantial lessening of competition in the UK?
So that's kind of the frame through which we've been looking at these deals.
Now, when I say looking at them, what that tends to mean is we have a phase one process,which is really a sort of information and evidence gathering process that gives the
(26:25):
parties themselves the opportunity to submit documents.
And they do a lot.
But also third parties, know, anyone can kind of submit evidence to this process.
It will all be looked at in the round by experts within the CMA.
So we've been running through this handful of cases in a, I think, a very timely andefficient way, looking at that evidence.
(26:49):
And as we go, we've been really striving to give greater clarity.
You know, we've been ourselves kind of hoping to get greater clarity around the sorts ofdeals that do and don't meet the two
criteria that I spoke about earlier.
And also, we've been publishing very fulsome decisions.
Not every authority around the world publishes a full decision if it's a clearance, right?
(27:12):
We do do that.
So we've been really trying to unpack our thinking in there because we know how importantit is for businesses, right, to have that clarity, especially investors.
So we've been really trying to do that.
And, and we've been actually publishing supplementary commentary on top just to just totry and
just to try and be as transparent as we possibly can.
(27:35):
And that's really the way that the phase one process is designed to work here in the UK.
It's a little unfamiliar perhaps for folks across the pond.
The other thing I would say just to put all of that into context is there are around50,000 &A deals announced every year.
Here in the UK, we took a look last year at 913 of those.
(27:58):
54 we investigated.
And one was blocked.
Just one.
So I think, you know, it's very fair to say that the way that we treat this is consciouslyproportionate.
That's the way the legislation is designed.
There's a very, very high bar.
And that's certainly our attitude when we look at these things.
(28:20):
What do you see as the major open issues or open questions right now for AI?
So, you we've talked about the competition piece quite a lot.
So let me sort of move away from that and talk about another couple of issues that Ithink, you know, should be on everybody's radar.
(28:41):
So the first one of those is the other half of our remit, which is consumer protection.
You know, this is this is very much in everyone's minds.
I'm sure that it is with yours, given the name of your podcast.
As I say, algorithmic harms have been, you know, something on our minds for some time.
And, you know, the report that we put out in April also looked quite in depth at thepropensity of AI to exacerbate those harms.
(29:05):
Now, you know, we also know that AI has incredible potential to do fantastic
things, right, for consumers.
But I think we want to be thoughtful about those two principles I mentioned of the CMAs,transparency and accountability, particularly in this context.
I think it's incredibly important.
(29:26):
Secondly, I think AI safety is a big one, right, that, you know, we should talk moreabout.
This is one which is kind of, you know,
catapulted up the political agenda for good reasons in the last couple of years.
And I think it's extremely important that we maintain the global momentum that we have nowbehind this.
(29:51):
And then the other piece I would sort of add on a little bit is what I talked about at thebeginning in terms of all of the different areas of regulation that need to work together
here.
So it's quite easy, I think, for all of us in AI to lose sight of the bigger picture
But you hear a lot at the moment around energy, for example, and the importance of thatwithin the AI ecosystem.
(30:13):
We know that it's incredibly important that infrastructure and planning work in the rightway in order to facilitate the growth of these industries.
So there's a big suite of policy issues there.
Absolutely.
I'm going to skip the question about coordinating with other agencies just for time, ifthat's okay.
Yeah, so we're just going.
So I'll do the DMCCA one next.
(30:42):
You mentioned earlier the DMCCA, the UK Digital Markets Competition and Consumers Act,which is different from the Digital Markets Act, the DMA in the European Union.
So could you say a little bit more about how that legislation impacts your authority andability to address these kinds of issues?
(31:04):
It is different, it is different.
And I think it's worth noting that, you know, whilst
authorities all over the world are taking different approaches to this.
Many of the principles are the same, right?
I think there's been a broad acknowledgement that, you know, for an innovation-led,fast-paced markets like digital, the sort of regulation that we had for quite a long time
(31:27):
just wasn't going to cut it, right?
So different places around the world have been developing their own kind of responses tothat.
And the digital markets competition and consumer regime in the UK is ours.
We like to
think that it is a very, very well designed and unique piece of regulation.
There's a few features which make it so, I'll talk to you about those again.
(31:51):
coming back to the principle I described earlier about how competition can unleash andunlock growth and investment, this is also the basis for this piece of regulation.
The idea here is to maximize the opportunities to grow a UK tech sector.
All right.
And that's incredibly exciting.
(32:12):
So that should mean that opportunities for the largest companies remain, right?
They play an incredibly important role in these digital markets.
And, you know, that should remain, but it means that the CMA will also have an ability toprotect and enhance opportunities for the smaller players, right?
For those scrappy startups, for the challenges, right?
(32:33):
To come in and have that fair shot that I talked about earlier.
And it should also mean fantastic opportunities for consumers and business customers toget great choices and fair deals.
So that's kind of the framing context for this.
We're hoping it's going to be genuinely a new era in digital growth for the UK economy.
(32:53):
So how is it going to do that exactly?
So it is supposed to apply to only a very small number of the very largest firms.
So those firms will have what's called strategic market status.
So this is a bit legalistic, but effectively what it means is they will have a substantialand entrenched position in one or more digital markets in the UK.
(33:20):
Now we haven't made any provisional decisions yet about where we're to focus thoseinvestigations.
But what I can talk about a bit is the principles and the process underpinning the waythat this regulation will work and why we think it's so special and why we think it should
be extremely attractive actually to
to businesses and investors in the way that it operates.
(33:40):
So first of all, it's designed to be very tailored and bespoke.
It's not a big sweeping, one size fits all piece of regulation.
We think that's really, really important for the reasons that I've described earlier.
Instead, what it allows us to do is really target specific homes with tailored solutions,working iteratively with the companies in a participative conversation that involves the
(34:05):
whole ecosystem, a wide range of stakeholders, so that the interventions we're makingreally are targeted to exactly that problem.
And it allows us to be quite iterative and bespoke in the way that we're doing it and veryproportionate as well.
This is a truly different approach to what I think we see in other pieces of equivalentregulation.
(34:27):
And certainly, this is something that we think is going to be very attractive, actually,to businesses.
And certainly, we hope to the consumers that will also benefit from all of this.
Now, AI is a consideration within that.
We've kind of talked about that a little bit.
How that might work in practice, for example, is where AI is part of one of the criticalinputs that we might look at in one of those digital activities.
(34:57):
cloud, for example.
Equally, in that deployment layer that I said I'd come back to when you're looking atthings like mobile ecosystems, search, productivity software, these are all potential
avenues for foundation models to be deployed and released across the economy.
That is also something that we can look at through the lens of this regulation.
(35:21):
Would you say that you've learned anything even before the DMCCA?
Yeah, that's okay.
That's okay.
While you're doing that, I'm just going to check something very, very quickly.
Yeah.
then we're almost done.
Just two more questions.
(35:42):
Shoot, you know, I misspoke and I knew I misspoke when I said it.
Yeah, I tell you what I'm going to do super quick.
I'm going to just say at the end of what I was talking about mergers, I said, I'm going tostep back for some context here.
There are 50,000 deals, right?
I'm just going to do that one little bit again if we can.
(36:03):
Thank you so much.
Damn lies and statistics.
Right.
Okay, let me just reverse for a second.
Yes.
Okay.
I think it's also worth just sort of noting for context here that around 50,000 &A dealswere announced last year.
(36:26):
And the CMA looked at 913.
Now 95 % of those didn't even proceed to an investigation, let alone an intervention ofsome kind.
And in fact, just one was prohibited.
So it's quite useful to keep those sort of figures in mind.
(36:46):
Super, thank you.
All right.
And how do you want to, where do you want to go next?
So I was just gonna ask you, what have you learned from, even pre-DMCCA, what have youlearned from the experience trying to exercise oversight over the big platforms?
God.
If you don't, I mean if that doesn't sound right, we could, you know, we've got enoughtime.
be sort of slightly encroaching on Will's territory there.
(37:10):
That's fine.
No, don't worry about it.
We've got enough content.
So why don't I just go on to the last one, which is the global piece.
That's totally fine.
Okay.
Because yeah, no, we've got, mean, there's, this is more than enough time even without anythings.
Great, okay.
So last question, the UK hosted a major high level AI safety summit in late 2023 thatresulted in an international declaration signed by I think 28 different countries around
(37:42):
AI policy.
To what extent do you think there is potential going forward for truly global governanceof this phenomenon given how global the technology is?
Very good question.
makes me feel a little tired when you ask it like that.
In my previous job, I had a big glass wall.
(38:04):
And on that glass wall, every time I found a new AI ethics or governance framework, Iwould just note it down and I ran out of wall really quite quickly.
So, you when you think about kind of global harmonization, yes, it seems it seems a longway off and to an extent that there are good reasons for that, right?
know, nations are sovereign in terms of the, you know, the legal frameworks that they haveand the decisions that they take, you know.
(38:30):
different countries are at different stages of technological development, have differentsocial norms.
So, you know, that I think is understandable and healthy.
I think when you when you think about an issue like this, and you look at the sort ofexistential risks that were discussed at Bletchley and in those sorts of conversations, it
(38:50):
is eminently important that we do have that sort of global consensus and globalconversation.
I think Bletchley was incredibly important in that.
it's fantastic that that
is continuing around the world.
from a competition perspective, we're a really active part of our international communitytalking about these things.
know, groups like the G7 just issued a communique last month on a commitment in this area.
(39:18):
Equally, we put out a joint statement over the summer with the FTC, the DOJ and theEuropean Commission around upholding competition and consumer protection in the AI space.
So we can all be we can all be doing that and, you know, continuing to to kind of
share best practice to support one another in terms of understanding and sharing the kindof capabilities and expertise that you need in this space.
(39:43):
These are borderless markets, right?
And we live in a very globally interconnected world.
So those relationships are strong already.
They've been getting stronger over the last several decades.
And I think they will continue to strengthen around an issue like AI.
I mean, it's interesting, we were talking earlier about the different approaches even todigital regulation being taken around the world.
(40:08):
In the EU now they have a big AI act.
In the US, I know it's kind of a bit of a fragmented picture.
Last I heard California's got 32 AI laws of its own.
And here in the UK, I say, we're, CMA is gonna take this principles-based approach.
Government's approach is kind of look at this through the lens of sectorial regulationprimarily, and then it has
(40:30):
the AI safety work as a kind of different strand and it will be bringing forward an AIsafety bill within this parliament.
So all of these different approaches, I hope, add up to some alignment on the need forreal action in this space.
But true harmonization, think we may need to come back in a few years and do anotherpodcast on that.
(40:53):
Well, that's all right.
Covered a lot and this is a very fast moving area.
Jessica, really appreciate your taking the time to speak with me.
Not at all.
Thanks so much, Kevin.
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.