Conference of February 14th 2022, 10h30 -12h.
On 14th February 2022, the Brussels Data Privacy Hub and EDHEC Business School co-hosted the
Conference on Data Sustainability, following last year’s edition. This year aimed to look at data
sustainability through the prism of Artificial Intelligence (AI). Speakers were Ben Wagner (TU Delft &
Sustainable Media Lab, Inholland), Christian D’Cunha (European Commission, DG Connect,
Cybersecurity and Digital Privacy Unit) and Ivana Bartoletti (Wipro, University of Oxford). Alessandra
Calvi (Brussels Privacy Hub, VUB, CYU) moderated the discussion.
The conference started with an introduction by Gianclaudio Malgieri (Brussels Privacy Hub, VUB,
EDHEC). He first addressed the definition of the word “sustainability”, whose etymology roots in the
Latin word sustinere, which means to hold up, to bear, to tolerate, to exist despite something. About
data privacy, this brings up the question of data privacy current systems’ ability to exist in the future
and under what conditions.
He highlighted that, just like environmental matters, data privacy should be addressed at a collective
scale, and not individually. “Digital information is the fuel of the new economy. But like the old
economy’s carbon fuel, it also pollutes” (Data pollution, by Omri Ben-Shahar). This marks the
introduction of the comparison between data sustainability and environmental sustainability.
Whereas the traditional sustainability scheme builds upon the triangle Economic-Social-Environment,
G. Malgieri suggested that “Data” substitutes “Environment”.
Going on with the parallel between environment and data, he highlighted that both address a problem
of negative externalities (private and collective), and they also use similar tools (impact assessment,
certifications…).
There is more to data sustainability than the environmental aspects
B. Wagner stressed that despite data privacy professionals are trying to reflect on the connection
between data privacy and sustainability, there are no obvious answers to this issue. Too often, people
associate sustainability only with the environment, whereas there is more to it.
C. D’Cunha added that lately discussions on AI have replaced Big Data in the public debate, but these
two issues are substantially the same. They are both resource-intensive, have an impact on the
environment and also on people.
He questioned to what extent the current path followed by governments is sustainable, considering
that a truly sustainable approach would address at the same time social justice, human rights and the
environment. He also put forward the need for a sense of scale, as the problems of data processing
are not equally distributed on the planet.
I. Bartoletti pointed out that the discussions on sustainability have multiplied both at a policy and
business level. Yet, this bears the risk of greenwashing and ethics washing. Avoiding falling into this
trap would require a structural shift into our approach to technology, less technocentric and more
collective.
Data processing: both a positive and a negative impact on the environment
B. Wagner noted that the simplification of our lives through technologies, in turn, complexifies data
processing and requires more energy. Software and data processing are getting all more costly in terms
of energy. He pointed out a tendency to develop media systems that are less sustainable than the
previous ones. The only way to tackle this issue seems to be their limitation.
He also suggested regulating at a European Union (EU) level the emissions of data centers. Indeed,
whereas the EU appears to favor the implementation of local data centers due to privacy reasons, data
storage is particularly pollutant in Europe.
I. Bartoletti noted that even if AI might help solve certain environmental issues, it is not a panacea. AI
has a terrible impact on the environment. She suggested that a proportionality framework should be
considered in the current AI regulation proposal. Also, the task performed by an AI solution should be
coherent with its ecological footprint.
She emphasized that this topic should be addressed at a political scale and the choice of the more
ecological options should not solely be left to the consumer. She noted how the lack of information on
AI’s environmental impacts depends on the lack of incentives for companies to release any information
on their global environmental impacts. Thus, a first step would be a clear political recognition of the
damage AI can cause to the environment.
There is a public interest in improving the data processing by AI
B. Wagner spotted a difficulty in addressing both collective rights and individual rights under existing
legal frameworks. He underlined that the interpretation of rights is rather individualized now, leaving
a limited place for collective challenges. He doubted that a market-based approach involving the
creation of costs related to unsustainable practices would be sufficient to create an individual incentive
for sustainable choices. He underlined that engaging with users is quite difficult because people need
concrete tradeoffs to accept more sustainable solutions.
C. D’Cunha called attention to the fact that AI is about inputs and outputs: data is the key element in
the use of AI. He thinks such data processing should be used to serve mankind. However, the General
Data Protection Regulation (GDPR) is not easily leveraged to assert collective rights.
Whereas the potential for AI would be to track non-personal subjects as pollution, nowadays the
tendency is to believe that AI’s stake is decoding and recoding the human brains and manipulating
them. The fact that facial recognition is a central topic of the AI act proposal demonstrates a lack of
accuracy in priorities.
On regulation efficiency, he also highlighted that the data protection authorities want to prohibit some
uses of data rather than impose sanctions, especially given that for large companies a sanction does
not have an impact significant enough to incent a change of practices. He also pointed out that
international action is unlikely to happen because of the fragmentation of the approaches to data
around the world.
On this, I. Bartoletti stated that the EU has a very strong counter push against the rush to data
collection. Indeed, there is no need for a gigantic amount of data but rather for smart-sized datacentric solutions to big issues in AI. She also thinks that the AI process should be challenged more:
what does AI do? Does it improve the well-being of the population? What is its impact on social,
economic, legal aspects? Can we have a proportionality test?
She suggested that human intelligence and virtual intelligence should be combined more. Human input
in the AI process would improve the quality of the data and a more meaningful human intervention
could be useful.
Environment and data privacy defenders should realize they should merge their fights
B. Wagner observed that the context in data privacy and environment is similar as both branches suffer
the influence of powerful corporations willing to curb change on a policy and governance aspect. He
suggested that users question their trust in online platforms in the same way as in the main oil
producers to reach a governance model that could be appropriate both from an environmental and a
data privacy perspective.
C. D’Cunha’s reaction to this was that a way must be found to connect the environmental impact of
data practices with individual privacy and suggested that the principle of data minimization could be
helpful to do so. He witnessed that there is no real interaction between data protection authorities
and environmental groups for now but this is going to be imperative soon.
AI is an integral part of programmatic advertising. The entire ecosystem of online advertising generates
substantial amounts of data and distribution of data to intermediaries with a significant cost to privacy.
In this regard, the Belgian Data Protection Authority asserted that the existing IAB Cookie consent
framework is illegal under GDPR. C. D’Cunha underlined that this system is also very wasteful,
considering that this industry is responsible for the emission of 160 million tons of CO2. He raised the
question of the acceptability of such a situation, given the environmental and privacy costs.
Concerning the principle of data minimization, the GDPR does not stop the processing of data, but it is
supposed to give controllers the reflex of evaluating the impact of their actions before taking them
and avoid using personal data when it is not needed. And if that principle was effective, the carbon
footprint would be much smaller.