Second European AI Alliance

Second European AI Alliance Assembly overview

The second European AI Alliance Assembly was hosted online due to the COVID-19 pandemic, on Friday 9th October.

The second edition of the European AI Alliance Assembly took place last Friday 9th in a full day event which was hosted online due to the COVID-19 pandemic. The Assembly gathered together more than 1,400 viewers who followed the sessions live and were also given the option to submit their questions to panellists.

The event

This year’s edition had a particular focus on the European initiative to build an Ecosystem of Excellence and Trust in Artificial Intelligence. The sessions were broken into plenaries, parallel workshops and breakout sessions. 

The following topics were addressed:

As a member of the European AI Alliance, Aphaia was pleased to join the event and enjoy some of the sessions addressing topics which are crucial in the development and implementation of AI in Europe, such as “Requirements for trustworthy AI” and “AI and liability”.

Requirements for trustworthy AI

The speakers shared their views on the risks brought by AI systems and the approaches that should be taken to enable the widespread use of AI in the society.

Hugues Bersini, Computer Science Professor at Free University of Brussels, considered that there is a cost function whenever AI is used, and optimizing it is the actual goal: “Whenever you can align social cost with individual cost there are no real issues”.

Haydn Belfield, Academic Project Manager at CSER Cambridge University, claimed that the high risk AI systems may imply for people life chances and their fundamental rights demand a framework of regulation including mandatory requirements that should, at the same time, be  flexible, adaptable and practical. 

For Francesca Rossi, IBM fellow and the IBM AI Ethics Global Leader, transparency and explainability are key. She explained that AI should be used to support the decision making capabilities of human beings, who have to make informed decisions. This purpose cannot be achieved if AI systems are a black box.

As response to the audience questions, the speakers discussed together how many risk levels would be necessary for AI. The main conclusion was that considering that only defining high risk is already a challenge, having two risk levels (high risk and not high risk) would be a good start on which further developments may be built in the future.

The speakers briefly talked about each of the requirements highlighted by the AI-HLEG for trustworthy AI, namely: human agency and oversight, technical robustness and safety, privacy and data governance, transparency, diversity, non-discrimination and fairness, societal and environmental well-being and accountability. 

In our view the discussions on AI and biases and human oversight where especially relevant:

AI and biases

Paul Lukowicz, Scientific Director and Head of the Research Unit “Embedded Intelligence” at the German Research Center for Artificial Intelligence defined machine learning as giving the computer methods with which it can extract procedures and information from data and stated that it is the core of the current success of AI. The challenge is that a lot of biases and discrimination in AI system come from giving data in which there are biases and discrimination. The challenge is that it is not that the developers somehow fail in giving data which is not per se representative: they actually use data that is representative and because there are discrimination and bias in our everyday life, this is what the systems learn and empathize. Linked to this issue, he considers that another pitfall is  the uncertainty, as there is no data set that covers the entire world. “We always have a level of uncertainty in life, so we have in AI systems”. 

Human oversight

Aimee Van Wynsberghe, Associate Professor in Ethics and Technology at TU Delft, raised some obstacles to human oversight:

  1. She challenged the fact that the output of an AI system is not valid until it has been reviewed and validated by an human. In her view, this can be quite difficult because there are biases that threaten human autonomy: automation bias, simulation bias, confirmation bias. Humans have a tendency to favor suggestions from automated decision-making system and ignore contradictory information that is made without automation. The other challenge in this regard is that having an AI system creating output and the human overviewing and validating is very time and resources consuming.
  2. As for the alternative based on the fact that the outputs of the AI system would become immediately effective but only if human overview is ensured afterwards, Aimee pointed out the issue of allocating the responsibility of ensuring human intervention:  “Who is going to ensure that human intervention happens? The company? Is the customer who would approach the company otherwise? Is it fair to assume that customers would have the time, the knowledge and the ability to do this?
  3. The monitoring of the AI system while in operation and the ability to intervene in real time and deactivate it would be difficult too because of human psychology: “there is lack of situational awareness that does not allow for the ability to take over”.

AI an liability

Corinna Schulze, Director of EU Government Affairs at SAP; Marco Bona, PEOPIL’s General Board Member for Italy and International Personal Injury Expert; Bernhard Koch, Professor of Civil and Comparative Law and Member of the New Technologies Formation of the EU Expert Group on liability for new technologies; Jean-Sébastien Borghetti, Private Law Professor at Université Paris II Panthéon-Assas and Dirk Staudenmaier, Head of Unit and Contract Law in the Department of Justice of the European Commission discussed the most important shortcomings on AI liability and put it in connection with the Product Liability Directive.

The following issues of the Directive were pointed out by the experts:

  • Time limit of 10 years: in the view of most of the speakers this may be an issue because it concerns producers only, which could be difficult whenever operators, users, owners and other stakeholders are involved. Furthermore 10 years is fine with  traditional products but it may not work in terms of protection of victims in relation to some AI artefacts and systems.
  • Scope: it concerns the protection of the consumers while it not address the protection of victims. Consumers and victims sometimes overlap but not always.
  • Notion of defect: it may cause some troubles with the distinction between product and services. The Directive covers only product, not services which may rise some concerns in relation to internet of things and software.

 

The Commission has made available the links of the sessions for all those who did not manage to attend the event or would like to see one or more session again.

Do you need assistance with AI Ethics? We can help you. Our comprehensive services cover both Data Protection Impact Assessments and AI Ethics Assessments.

EDPB Guidelines on the targeting of social media users overview

On 2nd September, the EDPB adopted their Guidelines 8/2020 on the targeting of social media users, which aim to clarify the implications that these practices may have on privacy and data protection.

Most social media platforms allow their users to manage their privacy preferences by enabling the option to make their profiles public or private. Pictures, videos and text are not the only personal information processed in this context though: what about analytics used to target social media users? Analytics are also personal data and they should be managed and protected accordingly. The European Data Protection Board (EDPB) is aware of the risks this creates to the fundamental rights and freedoms of individuals and has published these guidelines to provide their recommendations with regard to the roles and responsibilities of targeters and social media providers.

Actors involved in social media targeting

The EDPB explains the concepts of social media providers, users and targeters as follows:

  • Social media providers should be understood as providers of online platforms that enable the development of networks and communities of users, among which information and content is shared.
  • Users are the individuals who are registered with the service and create accounts and profiles which data is used for targeting purposes. This term also comprises those individuals that access the services without having registered.
  • Targeters are defined as natural or legal persons that communicate specific messages to the users of social media in order to advance commercial, political, or other interests, on the basis of specific parameters or criteria.
  • Other actors who may be also relevant are marketing service providers, ad networks, ad exchanges, demand-side and supply-side platforms, data brokers, data management providers (DMPs) and data analytics companies.

Identifying the roles and responsibilities of the various actors correctly is key in the process, as the interaction between social media providers and other actors may give rise to joint responsibilities under the GDPR.

Risks to the rights and freedoms of users

The EDPB highlights some of the main risks that may be derived from social media targeting:

  • Uses of personal data that go against or beyond individuals’ reasonable expectations.
  • Combination of personal data from different sources.
  • Existence of profiling activities connected to targeting.
  • Obstacles to the individual’s ability to exercise control over his or her personal data.
  • Lack of transparency regarding the role of the different actors and the processing operations.
  • Possibility of discrimination and exclusion.
  • Potential possible manipulation of users and undue influence over them.
  • Political and ideological polarisation.
  • Information overload.
  • Manipulation over children’s autonomy and their right to development.
  • Concentration in the markets of social media and targeting.

Relevant case law

The EDPB analyses the respective roles and responsibilities of social media providers and targeters through the relevant case law of the CJEU, namely the judgments in Wirtschaftsakademie (C-210/16) and Fashion ID (C-40/17):

– In Wirtschaftsakademie, the CJEU decided that the administrator of a so-called “fan page” on Facebook must be regarded as taking part in the determination of the purposes and means of the processing of personal data. The reasoning behind this decision is that the creation of a fan page involves the definition of parameters by the administrator, which has an influence on the processing of personal data for the purpose of producing statistics based on visits to the fan page, using the filters provided by Facebook.

– In Fashion ID, the CJEU decided that a website operator can be a considered a controller when it embeds a Facebook social plugin on its website that causes the browser of a visitor to transmit personal data of the visitor to Facebook. However, the liability of the website operator will be “limited to the operation or set of operations involving the processing of personal data in respect of which it actually determines the purposes and means”, therefore the website operator will not be a controller for subsequent operations carried out by Facebook after the data has been transmitted.

Roles and responsibilities of targeters and social media providers

Social media users may be targeted on the basis of provided, observed or inferred data, as well as a combination thereof.

In most cases both the targeter and the social media provider will participate in determining the purpose (e.g. to display a specific advertisement to a set of individuals social media users who make up the target audience) and means (e.g. by choosing to use the services offered by the social media provider and requesting it to target an audience based on certain criteria, on the one hand and by deciding which categories of data shall be processed, which targeting criteria shall be offered and who shall have access, on the other hand) of the processing personal data, therefore they will be deemed to be joint controllers pursuant to the Article 26 GDPR.

As pointed out by the CJEU in Fashion ID, the joint controllership status will only extend to those processing operations for which the targeter and the social media provider effectively co-determine the purposes and means, such as the processing of personal data resulting from the selection of the relevant targeting criteria, the display of the advertisement to the target audience and the processing of personal data undertaken by the social media provider to report to the targeter about the results of the targeting campaign. However, the joint control does not extend to operations involving the processing of personal data at other stages occurring before the selection of the relevant targeting criteria or after the targeting and reporting has been completed.

The EDPB also recalls that actual access to personal data is not a prerequisite for joint responsibility, thus the above analysis would remain the same even if the targeter only specified the parameters of its intended audience and did not have access to the personal data of the affected users.

Legal bases of the processing

It is important to note that, as joint controllers, both the social media provider and the targeter must be able to demonstrate the existence of a legal basis pursuant to the Article 6 GDPR to justify the processing of personal data for which each of the joint controllers is responsible.

In general terms, the two legal basis that are more likely to apply are legitimate interest and data subject’s consent.

In order to rely on legitimate interest as the lawful basis, there are three cumulative conditions that should be met:

– (i) the pursuit of a legitimate interest by each the data controller or by the third party or parties to whom the data are disclosed;
– (ii) the need to process personal data for the purposes of the legitimate interests pursued, and
– (iii) the condition that the fundamental rights and freedoms of the data subject whose data require protection do not take precedence.

In addition, opt-out should be enabled in a manner that data subjects should not only be provided with the possibility to object to the display of targeted advertising when accessing the platform, but also be provided with controls that ensure the underlying processing of his or her personal data for the targeting purpose no longer takes place after he or she has objected.

Legitimate interest will not be suitable in some circumstances though, therefore consent will be required in those cases. Intrusive profiling and tracking practices for marketing or advertising purposes that involve tracking individuals across multiple websites, locations, devices, services or data-brokering would be some of the examples.

The EDPB further notes that the consent collected for the implementation of tracking technologies needs to fulfil the conditions laid out in Article 7 GDPR in order to be valid. They highlight that pre-ticked check-boxes by the service provider which the user must then deselect to refuse his or her consent do not constitute valid consent. Moreover, based on recital 32, actions such as scrolling or swiping through a webpage or similar user activity would not under any circumstances satisfy the requirement of a clear and affirmative action, because such actions may be difficult to distinguish from other activity or interaction by a user, which means that determining that an unambiguous consent has been obtained would also not be possible. Furthermore, in such a case, it would be difficult to provide a way for the user to withdraw consent in a manner that is as easy as granting it.

The controller that should be in charge of collecting the consent from the data subjects will be the one that is involved first with them. This is because consent, in order to be valid, should be obtained prior to the processing. The EDPB also recalls that the controller gathering consent should name any other controllers to whom the data will be transferred and who wish to rely on the original consent.

Finally, where the profiling undertaken is likely to have a “similarly significant [effect]” on a data subject (for example, the display of online betting advertisements), Article 22 GDPR shall be applicable. An assessment in this regard will need to be conducted by the controller or joint controllers in each instance with reference to the specific facts of the targeting.

The EDPB welcomes comments to the Guidelines until 19th October.

You can learn more about joint controllership in our recent blog Joint controllership: key considerations by the EDPB.

 

Are you targeting social media users? You may need to adapt your processes to comply with the GDPR and the EDPB Guidelines. We can help you. Aphaia provides both GDPR, Data Protection Act 2018 and ePrivacy adaptation consultancy services, including data protection impact assessments, CCPA compliance and Data Protection Officer outsourcing.

joint controllership

Joint controllership: key considerations by the EDPB

The EDPB provides key considerations to clarify the concepts of processor, controller and joint controller in their Guidelines 07/2020.

The European Data Protection Board (EDPB) published their Guidelines 07/2020 on the concepts of controller and processor in the GDPR on 7th September, which aim to offer a precise meaning of these concepts and a criteria for their correct interpretation that is consistent throughout the European Economic Area.

Since the CJEU considered, in its Judgment in Fashion ID, C-40/17, the fashion retailer Fashion ID to be a controller jointly with Facebook by embedding the ‘Like’ button in its website, the concept of joint controllership seems to have a broader meaning, as it may apply now to some data processing that were deemed otherwise in the past.

In our blog today we go through the main insights provided by the EDPB with regard to the concept of joint controller.

The concept of joint controller in the GDPR

Pursuant to the Article 26 of the GDPR, the qualification as joint controller may arise where two or more controllers jointly determine the purposes and means of processing. The GDPR also states that the actors involved shall determine their respective responsibilities for compliance by means of an arrangement between them, whose essence shall be made available to the data subjects. However, the GDPR does not contain further provisions that specify the details around this type of processing, such as the definition of ‘jointly’ or the legal form of the arrangement.

Joint participation

The EDPB explains that joint participation can take the form of a common decision taken by the two or more actors involved in the processing or result from converging decisions by them. Thus in practice, joint participation can take several different forms and it does not require the same degree of involvement or equal responsibility by the controllers in each case.

  • Joint participation through common decision. It means deciding together and involves a common intention.
  • Joint participation through converging decisions. This one results from the case law of the CJEU on the concept of joint controllers. According to the GDPR, the requirements the decisions should meet to be considered as converging on purposes and means are the following:
    • They complement each other.
    • They are necessary for the processing to take place in such manner that they have a tangible impact on the determination of the purposes and means of the processing.

As a result, the question that should be contemplated to identify converging decisions would be along the lines of “Would the processing be possible without both parties’ participation in the sense that the processing by each party is inseparable?”.

The EDPB also highlights that the fact that one of the parties does not have access to personal data processed is not sufficient to exclude joint controllership.

 

Jointly determined purpose(s)

The EDPB considers that there are two scenarios under which the purpose pursued by two or more controllers may be deemed as jointly determined:

  • The entities involved in the same processing operation process such data for jointly defined purposes.
  • The entities involved pursue purposes which are closely linked or complementary. Such may be the case, for example, when there is a mutual benefit arising from the same processing operation, provided that each of the entities involved participates in the determination of the purposes and means of the relevant processing operation.

Jointly determined means

Joint controllership requires that two or more entities have exerted influence over the means of the processing. However, this does not mean that each entity involved needs in all cases to determine all of the means. There might be different circumstances which would qualify as joint controllership where the rest of requirements are met, even where the determination of the means is not equally shared between the parties, for example:

 

  • Different joint controllers define the means of the processing to a different extent, depending on who is effectively in a position to do so.
  • One of the entities involved provides the means of the processing and makes it available for personal data processing activities by other entities. The entity who decides to make use of those means so that personal data can be processed for a particular purpose also participates in the determination of the means of the processing. For example, the choice made by an entity to use for its own purposes a tool or other system developed by another entity, allowing the processing of personal data, will likely amount to a joint decision on the means of that processing by those entities.

 

Limits of joint controllership

The fact that several actors are involved in the same processing does not mean that they are necessarily acting as joint controllers of such processing. Not all kind of partnerships, cooperation or collaboration imply qualification of joint controllers as such qualification requires a case-by-case analysis of each processing at stake and the precise role of each entity with respect to each processing. The EDPB provides a non-exhaustive list of examples of situations where there is no joint controllership:

  • Preceding or subsequent operations: while two actors may be deemed joint controllers with regard to a specific data processing where the purpose and means of its operations are jointly determined, this does not affect the purposes and means of operations that precede or are subsequent in the chain of processing. In that case, the entity that decides alone should be considered as the sole controller of said preceding or subsequent operation.
  • Own purpose: the situation of joint controllers acting on the basis of converging decisions should be distinguished from the case of a processor, since the latter, while participating in the performance of a processing, does not process the data for its own purposes but carries out the processing on behalf of the controller.
  • Commercial benefit: the mere existence of a mutual benefit arising from a processing activity does not give rise to joint controllership. For example, if one of the entities involved is merely being paid for services rendered, it is acting as a processor rather than as a joint controller.

For instance, the use of a common data processing system or infrastructure will not in all cases lead to qualify the parties involved as joint controllers, in particular where the processing they carry out is separable and could be performed by one party without intervention from the other or where the provider is a processor in the absence of any purpose of its own. Another example would be the transmission of employee data to tax authorities.

Joint controller arrangement

Joint controllers should put in place a joint controller arrangement where they determine and agree on, in a transparent manner, their respective responsibilities for compliance with the GDPR. The following list of non-exhaustive tasks should be specified by means of said arrangement:

  • Response to data subjects requests exercised pursuant to the rights granted by the GDPR.
  • Transparency duties to provide the data subjects with the relevant information referred in Articles 13 and 14 GDPR.
  • Implementation of general data protection principles.
  • Legal basis of the processing.
  • Security measures.
  • Notification of a personal data breach to the supervisory authority and to the data subject.
  • Data Protection Impact Assessments.
  • The use of a processor.
  • Transfers of data to third countries.
  • Organisation of contact with data subjects and supervisory authorities.

The EDPB recommends documenting the relevant factors and the internal analysis carried out in order to allocate the different obligations. This analysis is part of the documentation under the accountability principle.

When it comes to the form of the arrangement, even if there is no legal requirement in the GDPR for a contract or other legal act, the EDPB recommends that such arrangement be made in the form of a binding document such as a contract or other legal binding act under EU or Member State law to which the controllers are subject.

The EDPB welcomes comments to the Guidelines until 19th October.

Do you require assistance with GDPR and Data Protection Act 2018 compliance? Aphaia provides both adaptation consultancy services, including data protection impact assessments, CCPA compliance and Data Protection Officer outsourcing.

EU-US Privacy Shield

EU-US Privacy Shield invalidation business implications follow-up

Since the Court of Justice of the European Union (CJEU) invalidated the EU-US Privacy Shield in their Schrems II judgement delivered two weeks ago, many questions have arisen around international data transfers to the US.

After the invalidation of the EU-US Privacy Shield by the CJEU two weeks ago, as reported by Aphaia, data transfers to the US require another valid safeguard or mechanism that provides an adequate level of data protection similar to the one granted by the GDPR.

European Data Protection Board guidelines

With the aim of clarifying the main issues derived from the invalidation of the EU-US Privacy Shield, the European Data Protection Board (EDPB) has published Frequently Asked Questions on the Schrems II judgement. These answers are expected to be developed and complemented along with further analysis, as the EDPB continues to examine and assess the CJEU decision.

In the document, the EDPB reminds that there is no grace period during which the EU-US Privacy Shield is still deemed a valid mechanisms to transfer personal data to the US, therefore businesses that were relying on this safeguard and that wish to keep on transferring data to the US should find another valid safeguard which ensures compliance with the level of protection essentially equivalent to that guaranteed within the EU by the GDPR.

What about Standard Contractual Clauses?

The CJEU considered the SCC validity depends on the ability of the data exporter and the recipient of the data to verify, prior to any transfer, and taking into account the specific circumstances, whether that level of protection can be respected in the US. This seems to be difficult though, because the Court found that US law (i.e., Section 702 FISA and EO 12333) does not ensure an essentially equivalent level of protection.

The data importer should inform the data exporter of any inability to comply with the SCCs and where necessary with any supplementary measures and the data exporter should carry out an assessment to ensure that US law does not impinge on the adequate level of protection, taking into account the circumstances of the transfer and the supplementary measures that could be put in place. The data exporter may contact the data importer to verify the legislation of its country and collaborate for the assessment. Where the result is not favourable, the transfer should be suspended. Otherwise the data exporter should notify the competent Supervisory Authority.

What about Binding Corporate Rules (BCRs)?

Given that the reason of invalidating the EU-US Privacy Shield was the degree of interference created by the US law, the CJEU judgement applies as well in the context of BCRs, since US law will also have primacy over this tool. Likewise before using SCCs, an assessment should be run by the data exporter and the competent Supervisory Authority should be reported where the result is not favourable and the data exporter plans to continue with the transfer.

What about derogations of Article 49 GDPR?

Article 49 GDPR comprises further conditions under which personal data can be transferred to a third-country in the absence of an adequacy decision and appropriate safeguards such as SCCs and BCRs, namely:

  • Consent. The CJEU points out that consent should be explicit, specific for the particular data transfer or set of transfers and informed. This element involves practical obstacles when it comes to businesses processing data from their customers, as this would imply, for instance, asking for all customers’ individual consent before storing their data on Sales Force.
  • Performance of a contract between the data subject and the controller. It is important to note that this only applies where the transfer is occasional and only for those that are objectively necessary for the performance of the contract.

What about third countries other than the US?

The CJEU has indicated that SCCs as a rule can still be used to transfer data to a third country, however the threshold set by the CJEU for transfers to the US applies for any third country, and the same goes for BCRs.

What should I do when it comes to processors transferring data to the US?

Pursuant to the EDPB FAQs, where no supplementary measures can be provided to ensure that US law does not impinge on the essentially equivalent level of protection as granted by the GDPR and if derogations under Article 49 GDPR do not apply, “the only solution is to negotiate an amendment or supplementary clause to your contract to forbid transfers to the US. Data should not only be stored but also administered elsewhere than in the US”.

What can we expect from the CJEU next?

The EDPB is currently analysing the CJEU judgment to determine the kind of supplementary measures that could be provided in addition to SCCs or BCRs, whether legal, technical or organisational measures.

ICO statement

The ICO is continuously updating their statement on the CJEU Schrems II judgement. The latest version so far dates 27th July and it confirms that EDPB FAQs still apply to UK controllers and processors. Until further guidance is provided by EU bodies and institutions, the ICO recommends to take stock of the international transfers businesses make and react promptly plus they claim that they will continue to apply a risk-based and proportionate approach in accordance with their Regulatory Action Policy.

Other European Data Protection Authorities’ statements

Some European data protection supervisory authorities have provided guidance in response to the CJEU Schrems II judgement. While most countries are still considering the implications of the decision, some other are warning about the risk of non-compliance and a few of them like Germany (particularly Berlin and Hamburg) and Netherlands have openly stated that transfers to the US are unlawful.

In general terms, the ones that are warning about the risks claim the following:

  • Data transfers to the U.S. are still possible, but require the implementation of additional safeguards.
  • The obligation to implement the requirements contained in the CJEU’s decision is both on the businesses and the data protection supervisory authorities.
  • Businesses are required to constantly monitor the level of protection in the data importer’s country
  • Businesses should run a previous assessment before transferring data to the US.

The data protection supervisory authority in Germany (Rhineland-Palatinate) has proposed a five-step assessment for businesses. We have prepared the diagram below which summarizes it:

Can the level of data protection required by the GDPR be respected in the US?

The CJEU considered that the requirements of US domestic law and, in particular, certain programmes enabling access by US public authorities to personal data transferred from the EU, result in limitations on the protection of personal data which do not satisfy GDPR requirements. Furthermore, the CJEU stated that US legislation does not gran data subjects actionable rights before the courts against the US authorities. 

In this context, it seems difficult that a company could be able to demonstrate that they can provide an adequate level of data protection to personal data transferred from the EU, because basically it would have to bypass US legislation.

Latest moves in the US Senate does not shed light in this issue, because the “Lawful Access to Encrypted Data Act” was introduced last month. It mandates service providers and device manufacturers to assist law enforcement with accessing encrypted data if assistance would aid in the execution of a lawfully obtained warrant.

Do you make international data transfers to third countries? Are you affected by Schrems II decision? We can help you. Aphaia provides both GDPR and Data Protection Act 2018 consultancy services, including data protection impact assessments, and Data Protection Officer outsourcing. We also offer CCPA compliance servicesContact us today.