Free beta access. 50 free AI tokens each week. Optional top-ups available. See what is included
Opening Word document…

GDPR does NOT preclude AI – standards are vague enough to allow AI without real limits

EPRS 20 - European Parliamentary Research Service. (“The impact of the General Data Protection Regulation (GDPR) on artificial intelligence,” Panel for the Future of Science and Technology, June 2020, https://www.europarl.europa.eu/RegData/etudes/STUD/2020/641530/EPRS_STU(2020)641530_EN.pdf ) PJW

AI is compatible with the GDPR AI is not explicitly mentioned in the GPDR, but many provisions in the GDPR are relevant to AI, and some are indeed challenged by the new ways of processing personal data that are enabled by AI. There is indeed a tension between the traditional data protection principles – purpose limitation, data minimisation, the special treatment of 'sensitive data', the limitation on automated decisions– and the full deployment of the power of AI and big data. The latter entails the collection of vast quantities of data concerning individuals and their social relations and processing such data for purposes that were not fully determined at the time of collection. However, there are ways to interpret, apply, and develop the data protection principles that are consistent with the beneficial uses of AI and big data. The requirement of purpose limitation can be understood in a way that is compatible with AI and big data, through a flexible application of the idea of compatibility, which allows for the reuse of personal data when this is not incompatible with the purposes for which the data were originally collected. Moreover, reuse for statistical purposes is assumed to be compatible, and thus would in general be admissible (unless it involves unacceptable risks for the data subject). The principle of data minimisation can also be understood in such a way as to allow for beneficial applications of AI. Minimisation may require, in some contexts, reducing the 'personality' of the available data,rather than the amount of such data, i.e., it may require reducing, through measures such as pseudonymisation, the ease with which the data can be connected to individuals. The possibility of re-identification should not entail that all re-identifiable data are considered personal data to be minimised. Ratherthe re-identification of data subjectsshould be considered as creation of new personal data, which should be subject to all applicable rules. Re-identification should indeed be strictly prohibited unless all conditions for the lawful collection of personal data are met, and it should be compatible with the purposes for which the data were originally collected and subsequently anonymised. The information requirements established by the GDPR can be met with regard to AI-based processing, even though the complexity of AI application has to be taken into account. The information made available to data subjectsshould enable themto understand the purpose of each AI-based processing and its limits, even without going into unnecessary technical details. The GDPR allows for inferences based on personal data, provided that appropriate safeguards are adopted. Profiling is in principle prohibited, but there are ample exceptions (contract, law or consent). Uncertainties exist concerning the extent to which an individual explanation should be provided to the data subject. It is also uncertain to what extent reasonableness criteria may apply to automated decisions. gconcerning privacy by design and by default, do not hinder the development of AI systems, if correctly designed and implemented, even though they may entail some additional costs. It needs to be clarified which AI applications present high risks and therefore require a preventive data protection assessment, and possibly the preventive involvement of data protection authorities. Finally, the possibility of using personal data for statistical purposes opens opportunities for the processing of personal data in ways that do not involve the inference of new personal data. Statistical processing requires security measures that are proportionate to the risks for the data subject, and which should include at least pseudonymisation. The GDPR prescriptions are often vague and open-ended The GDPR allows for the development of AI and big data applications that successfully balance data protection and other social and economic interests, but it provides limited guidance on how to achieve this goal. It indeed abounds in vague clauses and open standards, the application of which often requires balancing competing interests. In the case of AI/big data applications, the uncertainties are aggravated by the novelty of the technologies, their complexity and the broad scope of their individual and social effects. It is true that the principles of risk-prevention and accountability potentially direct the processing of personal data toward a 'positive sum' game, in which the advantages of the processing, when constrained by appropriate risk-mitigation measures, outweigh its possible disadvantages. Moreover these principles enable experimentation and learning, avoiding the over- and underinclusiveness issues involved in the applications of strict rules. However, by requiring controllers to rely on such principles, the GDPR offloads the task of establishing how to manage risk and find optimal solutions onto controllers, a task that may be challenging as well as costly. The stiff penalties for non-compliance, when combined with the uncertainty on the requirementsfor compliance, may constitute a novel risk, which, rather than incentivising the adoption of adequate compliance measure, may prevent small companies from engaging in new ventures. Thus, the successful application of GDPR to AI-application depends heavily on what guidance data protection bodies and other competent authorities will provide to controllers and data subjects. Appropriate guidance would diminish the cost of legal uncertainty and would direct companies – in particular small ones that mostly need such advice – to efficient and data protection-compliant solutions. Some policy indications The study concludes with the following indications on AI and the processing of personal data. • The GDPR generally provides meaningful indications for data protection in the context of AI applications. • The GDPR can be interpreted and applied in such a way that it does not substantially hinder the application of AI to personal data, and that it does not place EU companies at a disadvantage by comparison with non-European competitors. • Thus, the GDPR does not require major changes in order to address AI applications

Sign in to recut, update, or save this card.