Digital Markets Act
● The DMA’s scope should be re ned to ensure consistency and re ect the focus on gatekeepers.
The DMA risks being
over-inclusive in some respects and under-inclusive in others, and leaves open the possibility for overlapping and
con icting EU and national rules. These issues could be addressed with re nements to the dra . In pa icular:
○ A icle 1(6)
should be amended to preclude national rules from regulating substantial y the same practices as the
○ A icle 2(2)(g)
should be removed to avoid covering a broad range of non-gatekeeper services that fal within the
de nition of ‘cloud’
○ A icle 3(2)(a)
should be removed or replaced with a service-level turnover threshold to avoid unequal treatment
between pla orms of the same size and impo ance simply because of the revenues of their owners.
● The framework for the application of the substantive rules should be re ned to enable a genuine regulatory dialogue
and avoid unintended consequences.
We welcome the DMA’s stated aim to enable a regulatory dialogue with
gatekeepers. To enable a genuine regulatory dialogue and avoid unintended harmful consequences, however, the rules on
the application of the DMA’s substantive provisions require re nement. In pa icular:
The Commission should specify the obligations under A icle 6 that a given gatekeeper must implement before
these obligations become binding for the gatekeeper and exposing it to nes.
Gatekeepers should be given the chance to request an exemption from individual obligation beyond the narrow
limits set by A icle 8, if they can demonstrate legitimate and proporatione considerations that justify such an
A icle 10 should permit the Commission to narrow or remove obligations over time -- not only add to them.
● There are several proposals we suppo .
The EC’s proposal provides a number of sensible suggestions for types of
conduct that are typical y harmful and can be identi ed as such. The fol owing provisions are relevant:
○ A icle 5(b)
rightly proscribes the use of most favoured nation clauses.
○ A icle 5(c)
facilitates developers promoting their o ers to end users.
○ A icle 5(d)
ensures that business users can raise concerns with public authorities.
○ A icle 6(1)(c)
promotes users’ choice of apps while ensuring scope for protecting the service’s integrity.
○ A icle 6(1)(e)
ensures that user choice wil be free from technical incumbrances.
○ A icle 6(k)
requires fair and non-discriminatory conditions on business users’ access to app stores.
● Other proposed rules pursue impo ant objectives but require re nement.
Some of the behavioral rules would bene t
from fu her clari cation and re nement to ensure they are practicable, meet the DMA’s objectives and avoid unintended
adverse consequences. These rules include: A icles 5(g) and 6(1)(g)
(disclosure of pe ormance-related information and
prices paid/received by adve isers and publishers), A icle 6(1)(a)
(using business users’ data in competition with them),A icle 6(1)(b)
(uninstal ing preloaded apps), A icle 6(1)(d)
(equal treatment in ranking), A icle 6(1)(f)
(interoperability),A icle 6(1)(h)
(data po ability), A icle 6(1)(i)
(data generated by business users), and A icle 12
● A small number of proposals would be harmful.
Ce ain proposals in the DMA would not improve the contestability of
digital markets and carry substantial risks to innovation and the welfare of European businesses and consumers. The
fol owing provisions are relevant:
○ A icle 5(a)
addresses user consent for combinations of personal data, which area already regulated by the GDPR.
Di erent paral el rules on the permissible extent of data processing is liable to lead to confusion, unce ainty, and
○ A icle 6(1)(j)
addresses third pa y access to search data, which is likely to harm innovation and expose search
engines to manipulation.
● Ce ain additional ideas warrant consideration.
It is wo h considering other possible solutions and measures, including a
role for choice screens and data mobility systems
○ Choice screens
can be an e ective way to address concerns about behavioural biases, including ‘default’ bias. At
present, choice screens are shown only on ce ain pla orms, including Google’s Android operating system. They
could be applied to other digital pla orms too to increase consumers’ awareness of alternatives and facilitate easier
switching among services. Choice screens could be considered as a way to ensure compliance with A icles 6(1)(c)
○ Data mobility systems
can intensify competition and enhance choice in digital markets just as Open Banking has
done in the nancial sector. Regulators can build on the Data Transfer Project, which has developed much of the
necessary technical infrastructure for technology companies, possibly specifying pa icipation in the Project as a
measure to comply with A icle 6(1)(h)
● Rules should have a clear connection to the DMA’s metrics for success.
The DMA claims that its rules wil signi cantly
increase GDP, employment, sales, and consumer surplus. Any rules that ultimately become pa of the GDP should have a
clear connection to achieving these goals.
A icle 2: De nitions
A icle 2(2) Core pla orm services
“‘Core pla orm service’ means any of the fol owing: (a) online intermediation services; (b) online search engines; (c) online social
networking services; (d) video-sharing pla orm services; (e) number-independent interpersonal communication services; (f)
operating systems; (g) cloud computing services; (h) adve ising services, including any adve ising networks, adve ising
exchanges and any other adve ising intermediation services, provided by a provider of any of the core pla orm services listed in
points (a) to (g);
The DMA proposal makes clear that it is concerned with pla orms that connect businesses with consumers, and which can
regulate that connection: see
A icle 3(1)(b) and recitals 6 and 12.
Recital 13, however, creates ambiguity because it suggests that “end users
” within the meaning of the DMA can also be “business
Consistent with the DMA’s stated objective and scope, it would therefore be useful to clarify that mere inputs and B2B services
do not qualify as “core pla orm services
” under the DMA. This is consistent with the DMA’s objective of addressing concerns
relating to services that “directly intermediate between business users and end users
” (Recital 12, emphasis added).
Absent clari cation, there is a risk of confusion as to whether mere inputs or B2B services would be caught, despite not having
the characteristics of a gatekeeper that regulates businesses’ access to end users. Examples could include data storage
solutions for businesses or business analytics tools.
A icle 2(2)(h): adve ising services
Adve ising services are treated as core gatekeeper services only if they are “provided by a provider of any of the core pla orm
services listed in points (a) to (g)
We understand -- and suppo -- the DMA’s aims of ensuring transparency in online adve ising. Recital 42 refers to sector-wide
concerns and the “sheer complexity of modern day programmatic adve ising
However, the current language of A icle 2(2)(h) brings adve ising services within scope only if they are o ered by a provider of
other core pla orm services.
This undermines the DMA’s objective of ensuring transparency because it leaves many impo ant adve ising services out of
scope that would meet the DMA’s quantitative thresholds. It would also lead to unjusti ed unequal treatment since the question
of whether a pa icular adve ising service is a ‘gatekeeper’ is unrelated to what other services are provided by the same
A icles 2(2)(g) and 2(11): cloud computing services
Cloud computing means a digital service as de ned in point 19 of A icle 4 of Directive (EU) 2016/1148 of the European
Parliament and of the Council
The de nition of cloud computing services in the NIS Directive is, by design, broad because the Directive aims to address issues
of cybersecurity -- it does not limit itself to services that regulate businesses’ access to users.
The NIS Directive de nition therefore covers al digital services that “al ow access to a scalable and elastic pool of shareable
”. The Commission’s accompanying Communication states that this de nition includes “so ware as a
The DMA’s cross-reference to the NIS Directive therefore would bring a broad range of services under its scope that have none
of the characteristics of gatekeeper pla orms. And it would create unequal treatment between online services and their
on-premises equivalents. It would, for example, apply to online email, calendaring, productivity, and database services, while
on-premises competitors of such services would not be covered.
A icle 3(2)(a): group level turnover / market capitalisation thresholds
This provision formulates a quantitative threshold at the level of the “unde aking to which
” the core pla orm service belongs of
“an annual EEA turnover equal to or above EUR 6.5 bil ion in the last three nancial years, or [. .] average market capitalisation or
the equivalent fair market value of [. .] at least EUR 65 bil ion in the last nancial year, [. .]
A turnover-based threshold for a given service may be a reasonable proxy for the signi cance of that service. Formulating a
turnover or market capitalization threshold at the group level, however, is liable to lead to unjusti ed unequal treatment.
The same kind of service with the same number of users may or may not be covered by the DMA depending on whether the
unde aking that owns the service reaches the speci ed revenue or value threshold, even if that unde aking’s revenues are
unrelated to the service in question.
As such, the DMA risks being both underinclusive by not covering services that are in fact impo ant (i.e.
, services with a large
number of users, but belonging to unde aking low turnover would not meet the group-wide turnover threshold). At the same
time it risks being overinclusive by covering services without market signi cance (i.e.
, services that generate li le revenue, but
that belong to an unde aking with signi cant revenues from unrelated sources would be covered).
The fact that the threshold operates as a presumption does not resolve this concern, since the general criteria are broadly
de ned and meeting the thresholds has considerable legal consequences.
B. Application of the rules
A icle 1(6) - overlapping national measures
“This Regulation is without prejudice to the application of A icles 101 and 102 TFEU. It is also without prejudice to the application
of: national rules prohibiting anticompetitive agreements, decisions by associations of unde akings, conce ed practices and
abuses of dominant positions; national competition rules prohibiting other forms of unilateral conduct insofar as they are applied
to unde akings other than gatekeepers or amount to imposing additional obligations on gatekeepers
A icle 1(5) stipulates that “Member States shal not impose on gatekeepers fu her obligations by way of laws, regulations or
administrative action for the purpose of ensuring contestable and fair markets
.” Similarly, Recital 9 acknowledges that national
competition rules “should not a ect the obligations imposed on gatekeepers under this Regulation and their uniform and
e ective application in the internal market
At the same time, however, A . 1(6) reserves the possibility for Member States to apply their competition rules. Retaining the
possibility for national competition laws to regulate substantively the same practices as the DMA is liable to frustrate the DMA’s
goal of achieving a harmonised set of rules on the contestability of digital markets. For example, the latest amendment to the
German Act against Restrictions of Competition -- like the DMA -- addresses practices concerning self-preferencing, data usage
The need for consistency across the DMA and national competition laws is recognised in the Legislative Financial Statement
accompanying the DMA, which refers to “an emerging fragmentation of the regulatory landscape
”; Members States addressing
“pla orm related problems at national level
” (including through competition measures); and “national rules in response to the
problems associated to the conduct of gatekeepers in the digital sector
” including as regards “MFN clauses
” (a competition
issue) (paragraphs 1.5.2-1.5.3).
Therefore, permi ing more extensive national measures alongside the DMA wil exacerbate -- not resolve -- the problem of
fragmented rules across Member States.
A icle 3(8) - direct applicability of A icle 6 obligations
“The gatekeeper shal comply with the obligations laid down in A icles 5 and 6 within six months a er a core pla orm service has
been included in the list pursuant to paragraph 7 of this A icle.
A icle 25 - direct exposure to non-compliance decisions for A icle 6 obligations without speci cation
“The Commission shal adopt a non-compliance decision in accordance with the advisory procedure referred to in A icle 32(4)
where it nds that a gatekeeper does not comply with one or more of the fol owing:
(a) any of the obligations laid down in A icles 5 or 6;
(b) measures speci ed in a decision adopted pursuant to A icle 7(2)
A icle 26 - direct exposure to nes for A icle 6 obligations without speci cation
“In the decision pursuant to A icle 25, the Commission may impose on a gatekeeper nes not exceeding 10% of its total turnover
in the preceding nancial year where it nds that the gatekeeper, intentional y or negligently, fails to comply with:
(a) any of the obligations pursuant to A icles 5 and 6;
(b) the measures speci ed by the Commission pursuant to a decision under A icle 7(2);
A icle 3(8) suggests that the obligations set out in A icle 6 are directly applicable and subject to nes for non-compliance under
A icle 26 without fu her speci cation.
This is in tension with Recitals 33 and 58 which envisage a “regulatory dialogue with gatekeepers
” and fu her speci cation for
obligations that “require speci c implementing measures in order to ensure their e ectiveness and propo ionality
”. In line with
these recitals, the obligations listed under A icle 6 are identi ed as being susceptible to fu her speci cation.
The direct applicability under A icle 3(8), however, means that a gatekeeper wil have to comply with A icle 6 - and is exposed to
high nes and non-compliance decisions - even before the envisaged regulatory dialogue can take place that would specify what
a pa icular gatekeeper is meant to do in concrete terms to comply in an e ective and propo ionate manner. Moreover, the
threat of nes and non-compliance decisions is inconsistent with the envisaged possibility to seek guidance under A icle 7(7).
This approach is liable to lead to legal unce ainty, sti e technical development, and undermine the envisaged concept of a
To enable a genuine regulatory dialogue and safeguard legal ce ainty, a be er approach is to provide for speci cation of the
obligations under A icle 6 prior to their applicability.
This approach is re ected in the recent amendment to the German competition law and the UK proposals for ‘pro-competitive
A icle 8 - possibility to justify practices
The dra DMA, unlike the amendment to the German competition act and the UK proposal for digital codes of conduct, does not
provide for a general possibility for gatekeepers to justify their conduct. Instead, A icles 8 and 9 include only narrow and highly
restrictive exceptions to the obligations under A icles 5 and 6.
Given the wide scope of the DMA, the onerous nature of its obligations, and the complex technologies involved, there is a serious
risk that a rigid application -- without a possibility to justify exceptions -- wil have dispropo ionate and counter-productive
consequences. Rather than promote openness and innovation, a rigid application may sti e innovation, diminish competition,
and harm businesses and end users.
By contrast, providing for a possibility to justify exceptions would future-proof the DMA and give it the requisite exibility to
achieve its objectives while avoiding unintended harmful consequences. This approach would also be consistent with the
approach adopted under the amendment of the German competition act and the UK proposals for digital codes of conduct.
A icle 10 - updating lists of obligations
“The Commission is empowered to adopt delegated acts in accordance with A icle 34 to update the obligations laid down in
A icles 5 and 6 where, based on a market investigation pursuant to A icle 17, it has identi ed the need for new obligations
addressing practices that limit the contestability of core pla orm services or are unfair
A icle 10 grants the Commission the possibility to add new obligations to A icles 5 and 6, fol owing a market investigation.
Recital 78 suggests that this possibility serves the need to ensure the regulation is future-proof. For the same reason, the
Commission should have the possibility to sunset or narrow obligations listed in A icles 5 and 6 if -- in light of technical,
commercial, or other developments -- it considers that those obligations are no longer warranted.
Moreover, ce ain DMA provisions address policy objectives that are covered in other regulations, which creates a risk of
misalignment (and possibly con icting rules). For example, the GDPR pursues -- and seeks to strike a balance between -- the
objectives of (i) enhancing user protection over the data, and (i ) promoting the free ow of data within the internal market. That
balance risks being material y altered by provisions of the DMA that deal with data combinations, access and po ability.
Accordingly, there is a need for a mechanism to ensure regulatory alignment as between the DMA and other legislative measures
-- A icles 10 and 17 can provide a solution in this regard.
A icle 2 - De nition of services
The DMA does not explain under what circumstances functionality o ered by a gatekeeper constitutes a distinct service as
opposed to an element of the core pla orm service provided by the gatekeeper.
Several of the DMA’s provisions govern the relationship between a gatekeeper’s core pla orm service and other separate or
‘ancil ary’ services that the gatekeeper provides. The DMA, however, provides no guidance as to how a separate service is to be
distinguished from the core pla orm service. Existing case law on tying provides a basis for identifying when a separate ‘service’
(as opposed to a mere product feature) exists.
A icle 38(2) -- Evaluation mechanism
“The evaluations shal establish whether additional rules, including regarding the list of core pla orm services laid down in point 2
of A icle 2, the obligations laid down in A icles 5 and 6 and their enforcement, may be required to ensure that digital markets
across the Union are contestable and fair. Fol owing the evaluations, the Commission shal take appropriate measures, which may
include legislative proposals
Consistent with our comments in respect of A icle 10, any review mechanism ought to include the possibility to remove or
narrow obligations -- not only to add them.
Any evaluation mechanism should -- as far as possible -- use objective, quanti able measures to evaluate whether the current list
of obligations are propo ionate and warrant retention in the DMA. The Legislative Financial Statement sets out expected ene ts
of the DMA that can be measured and ought, therefore, to be addressed in the envisaged evaluation process. These include: (i) a
1.5% annual increase in GDP, (i ) 1-1.4 mil ion new jobs, (i i) increased sales via smal er pla orms, and (iv) a 13 bil ion euro consumer
C. Speci c provisions
A icle 5(a) - Combination of personal data
“refrain from combining personal data sourced from these core pla orm services with personal data from any other services
o ered by the gatekeeper or with personal data from third-pa y services, and from signing in end users to other services of the
gatekeeper in order to combine personal data, unless the end user has been presented with the speci c choice and provided
consent in the sense of Regulation (EU) 2016/679.
We suppo rigorous protection of personal data. The proper legal instrument for that protection is the GDPR, which already
provides for a high level of protection.
Introducing rules on the protection of personal data in the DMA is liable to give rise to tension with the GDPR and create legal
unce ainty. For example, the GDPR al ows combining data across services on grounds other than user consent, such as
legitimate interests or a contract.
This is for good reasons: requiring individual consent for every instance of cross service use is liable to be cumbersome,
unwieldy, and disruptive for users and services, and may hinder or delay innovative developments and ve ical integration that
bene t users and businesses.
The obligation in A icle 5(a), moreover, contradicts existing case law on data combination: in its Facebook
Bundeska el amt expressly found that Facebook’s combination of data from di erent internal services was legitimate and
bene cial, and only chal enged the combination of data derived from third-pa y services.
At the same time, concerns related to proper protection and use of personal data apply universal y, regardless of the nature and
impo ance of a pa icular service. There is therefore no good reason to diverge on the level of protection of personal data
depending on the type of service.
A icle 5(g) - disclosure of prices for adve ising services
“provide adve isers and publishers to which it supplies adve ising services, upon their request, with information concerning the
price paid by the adve iser and publisher, as wel as the amount or remuneration paid to the publisher, for the publishing of a
given ad and for each of the relevant adve ising services provided by the gatekeeper.
We suppo transparency on the terms of adve ising services and make clear to our adve iser and publisher customers the
prices they pay. We have also published our aggregated take-rates so that publishers and adve isers understand what share of
spend we retain.
A icle 5g, however, could be read to require gatekeepers not only to be transparent about the prices that a given counterpa y
(adve iser or publisher) pays, but also to disclose the prices that other
counterpa ies pay on a granular basis. For example, a
publisher would see not only the fees it pays to the gatekeeper for publisher-facing services it uses (as is the case today), it
would also see the fees paid by every adve iser to the gatekeeper for adve iser-facing services. Conversely, an adve iser
would be able to see the fees every publisher pays for publisher-facing services.
There are signi cant problems with this approach:
First, it is not necessary to al ow a counterpa y to make decisions on what adve ising service to use. For example, a publisher
wil want to select the publisher-facing ad service that generates the most publisher revenue. The publisher does not need to
know the prices that adve isers pay for the services they purchase to make this selection. If Product A generates EUR2 revenue
and Product B generates EUR1 of ad revenue for the publisher per ad - it is irrelevant to the publisher’s choice that the adve iser
behind the bid on Product A paid 20 cents for adversier services while the adve iser behind the bid on Product B paid 10 cents.
Second, business pa ners can su er commercial harm, including reduced competition, from mandating disclosure of the fees
they pay. For example, adve isers could su er higher costs and reduced competition from the fees they pay being disclosed to
publishers. Many publishers compete to sel inventory direct
to adve isers (i.e. outside of the intermediation services of the
gatekeeper) in addition to using intermediation services. Tel ing publishers the fees paid by adve isers for intermediation
services wil give publishers commercial y sensitive information on the cost of the adve isers’ “outside option” when negotiating
direct deals and therefore risks reduced competition.
Accordingly, the recent amendment to German competition legislation does not require similarly far reaching disclosures but
focuses on disclosure of su cient information to assess the value of the service (§19a ARC, paragraphs 2(6)).
These considerations warrant treating A icle 5(g) as an A icle 6 measure to be speci ed by the European Commission, based
on the German provision. This can most e ciently be achieved by consolidating A icles 5(g) and 6(1)(g) into a single provision.
As regards A icle 6(1)(g) speci cal y, we have two principal comments. First, the requirement to provide data for free contrasts
with A icle 6(1)(j), which permits charges based on FRAND rates. Second, the clause ought not to require the disclosure of data
that might damage the integrity of the system, consistent with A icle 6(1)(c).
A icle 6(1)(a) -- Using businesses’ data to compete with them
“refrain from using, in competition with business users, any data not publicly available, which is generated through activities by
those business users, including by the end users of these business users, of its core pla orm services or provided by those
business users of its core pla orm services or by the end users of these business users
We suppo the principle that gatekeepers should not use data they gather from business users to disto competition in favour
of gatekeepers’ own rival businesses.
The language of Recital 43, makes clear that the concern is with data “gained from transactions” with a business user and the
use of these data by the gatekeeper for “similar services
”. We consider that, for clarity and legal ce ainty, this language should
be re ected in A icle 6(1)(a) itself.
In addition, we note the principle in recital 45 that obligations concerning cloud computing “should not a ect the right of
gatekeepers to use aggregated data for providing ancil ary data analytics services
.” The same principle ought to apply to other
product areas too -- there is no reason for limiting it to Cloud.
A icle 6(1)(b) -- Uninstal ing preloaded apps
“al ow end users to un-instal any pre-instal ed so ware applications on its core pla orm service without prejudice to the
possibility for a gatekeeper to restrict such un-instal ation in relation to so ware applications that are essential for the functioning
of the operating system or of the device and which cannot technical y be o ered on a standalone basis by third-pa ies
Recital 46 explains that this provision aims to prevent a gatekeeper from “favour[ing] its own services or products on its core
pla orm service, to the detriment of the same or similar services that end users could obtain through third pa ies
.” Any rule on
users’ ability to uninstal apps should, therefore, require the same conditions to be applied to al preloaded apps whether they
belong to gatekeepers or third pa ies.
Moreover, while A icle 6(1)(b) o ers an exception for apps that are essential to the operating system or the device, it omits
services on which other preloaded or downloaded apps may rely. It also overlooks the possibility to al ow users to disable apps,
which amounts to the same outcome as uninstal ing from a user’s perspective, but avoids much of the technical complexity that
comes from ful y removing an app from a device.
The be er approach, in our view, is to apply the wording of Recital 47, which envisages the possibility for gatekeepers to
“implement propo ionate technical or contractual measures
” to safeguard integrity.
A icle 6(1)(d) - Ranking
“refrain from treating more favourably in ranking services and products o ered by the gatekeeper itself or by any third pa y
belonging to the same unde aking compared to similar services or products of third pa y and apply fair and non-discriminatory
conditions to such ranking
“the gatekeeper should not engage in any form of di erentiated or preferential treatment in ranking on the core pla orm service,
whether through legal, commercial or technical means, in favour of products or services it o ers itself or through a business user
which it controls. To ensure that this obligation is e ective, it should also be ensured that the conditions that apply to such ranking
are also general y fair. Ranking should in this context cover al forms of relative prominence, including display, rating, linking or
We suppo principles that ensure ranking is free from a i cial manipulation and provides users with relevant results.
Ranking is a complex task especial y for services that o er results across a wide spectrum of di erent types of queries and
information. A crucial element to provide useful and relevant results is for a service to have the ability to use di erent formats
and di erent algorithms for di erent results.
The High Cou of England and Wales found, for example, in Streetmap
that Google showing a map in response to queries for
locations was bene cial for users and had no demonstrable adverse impact. In Shopping
, the European Commission recognized
that it was bene cial for Google to show specialized results with di erent formats and algorithms.
Yet the language in Recital 49 could be read to suggest that A icle 6(1)(d) is meant to outlaw any di erentiated treatment of
results. Such a position recently forced Google to remove Shopping Units in Turkey with detrimental consequences for countless
Turkish merchants, which lost tra c and now pay higher prices for promoting their o ers, and harm to users for whom it is now
more di cult to nd relevant product o ers. It also runs counter to the explanation on “fairness” in Recital 57 which identi es as
unfair only unjusti ed di erentiation and dispropo ionate advantages. While Recital 57 discusses fairness principles in the
context of app stores, there is no good basis for formulating di erent fairness principles across di erent obligations under
A icle 6(1). In pa icular, as Recital 57 makes clear a categorical and absolute prohibition of di erentiation - as suggested by
Recital 49 - is incompatible with a properly understood fairness standard.
The ambiguity created by Rectial 49 is compounded by the fact that A icle 6(1)(d) uses three di erent overlapping terms (no
disfavoring, fair, non-discriminatory) to denote the requisite ranking standard, which is liable to create confusion and unce ainty.
It is critical to clarify that A icle 6(1)(d) does not preclude gatekeepers engaging in justi able di erentiation (e.g.
di erent types of results with di erent formats and algorithms, or showing paid results). Otherwise, search services would be
forced to undo years of innovation and their ability to show useful and relevant results would be undermined.
A icle 6(1)(f) - Interoperability
“al ow business users and providers of ancil ary services access to and interoperability with the same operating system, hardware
or so ware features that are available or used in the provision by the gatekeeper of any ancil ary services”
We are active proponents for open and interoperable systems. We developed Android, for example, to be an open and ful y
interoperable operating system and we work on numerous industry initiatives to promote interoperability.
As developers of interoperable systems, we know, however, that enabling interoperability can pose di cult questions and
tradeo s in terms of functionality, quality assurance, user safety, and intel ectual prope y, for example. Critical y, enabling
interoperability presents di erent chal enges for an operating system (which by design is meant to interoperate with third-pa y
products) and application level services.
In line with these considerations, Recital 52 suggests that A icle 6(1)(f) is limited to providers of operating systems. However,
this limitation is not clearly a iculated in the wording of A icle 6(1)(f) which in conjunction with the introductory language to
A icle 6(1) could be read as applying in an undi erentiated manner to al core pla orm services.
An overly broad and rigid interoperability requirement is liable to prevent or delay technical development and innovation. We
therefore recommend clarifying that the obligation under A icle 6(1)(f) applies only to operating systems and recognizes the
need for a reasonable balance in the provision of interoperability (including, per Recital 41, an exemption for conduct that leads to
an “improvement of end user o ering
A icle 6(1)(h) -- Data po ability
“provide e ective po ability of data generated through the activity of a business user or end user and shal , in pa icular, provide
tools for end users to facilitate the exercise of data po ability, in line with Regulation EU 2016/679, including by the provision of
continuous and real-time access
We have long-suppo ed consumers’ control over their data, and realised some time ago that data po ability made users value
our services more. Since 2007, our projects to enhance data po ability include the Data Liberation Front, Google Takeout, and
the Data Transfer Project.
However, it is not clear what “continuous and real time
” access means in this context, or whether it is possible as a technical
ma er for any given product. The reference to “e ective
” data po ability provides the Commission with exibility to determine
whether a pa icular data po ability tool is su cient in a given situation, whether data is provided su ciently quickly, or whether
fu her measures are required.
A icle 6(1)(i) -- Data generated by business users
“provide business users, or third pa ies authorised by a business user, free of charge, with e ective, high-quality, continuous and
real-time access and use of aggregated or non-aggregated data, that is provided for or generated in the context of the use of
the relevant core pla orm services by those business users and the end users engaging with the products or services provided
by those business users; for personal data, provide access and use only where directly connected with the use e ectuated by the
end user in respect of the products or services o ered by the relevant business user through the relevant core pla orm service,
and when the end user opts in to such sharing with a consent in the sense of the Regulation (EU) 2016/679
We have a number of concerns with A icle 6(1)(i).
First, the de nition of what constitutes business users’ data appears exceptional y broad. For example, Recital 54 refers to “data
” business users’ activity. Inferred data would typical y include proprietary commercial or technical insights that
require engineering work to produce and may be used to improve Google’s services. A requirement to disclose business secrets
of this kind would undermine innovation.
Second, as noted in respect of A icle 6(i)(h), it is not clear what “continuous and real time
” (or “high quality
”) access means in
this context, or whether it is possible as a technical ma er for any given product.
Third, the requirement to provide data for free contrasts with A icle 6(1)(j), which permits charges based on FRAND rates.
A icle 6(1)(j) - access to search data
“provide to any third pa y providers of online search engines, upon their request, with access on fair, reasonable and
non-discriminatory terms to ranking, query, click and view data in relation to free and paid search generated by end users on
online search engines of the gatekeeper, subject to anonymisation for the query, click and view data that constitutes personal
Recital 56 contemplates the disclosure of “aggregated datasets containing information about what users searched for, and how
they interacted with, the results that they were served
”. A icle 6(1)(j), however, seems to go fu her by requiring disclosure of
ranking, query, click, and view data, subject only to anonymization.
We question the premise of this far reaching and intrusive disclosure and sharing requirement. There is no empirical evidence
that access to data is a material limiting factor in search engine competition. Nor is there a good basis for singling out search
data for regulatory disclosure and sharing obligations, as opposed to other data sets.
The disclosure obligation foreseen in A icle 6(1)(j) con icts with fundamental legal principles, including the protection of
intel ectual prope y, propo ionality, and equal treatment. We therefore suggest that this provision should be removed. In fact,
the disclosure and sharing obligation imposed under A icle 6(1)(j) contravenes established case law on the duty to share assets
Moreover, anonymization alone does not necessarily protect against privacy violations (see
a study in Nature by Yves Alexandre
one of the author’s of the Commission’s repo into ‘Competition Policy for the Digital Era’). And disclosure of
search data may have other serious, harmful consequences including exposing a search service to ranking manipulation and
copying of its results and search algorithms.
Should a disclosure obligation for search data, nonetheless be maintained, it must give due consideration to legitimate interests
standing against such disclosure.
A icle 12 -- Repo ing concentrations
“A gatekeeper shal inform the Commission of any intended concentration within the meaning of A icle 3 of Regulation (EC) No
139/2004 involving another provider of core pla orm services or of any other services provided in the digital sector irrespective
of whether it is noti able to a Union competition authority under Regulation (EC) No 139/2004 or to a competent national
competition authority under national merger rules”
We recognise the Commission’s interest in being made aware of transactions involving core pla orm services. However, the
requirement to repo transactions that do not involve core pla orm services is unrelated to the objectives of the DMA --
keeping markets contestable -- and undermines the DMA’s logic of applying obligations only to in-scope core pla orm services.