Privacy & Information Security Law Blog

On December 1, 2017, the Centre for Information Policy Leadership (“CIPL”) at Hunton & Williams LLP submitted formal comments to the Article 29 Working Party (the “Working Party”) on its Guidelines on Automated Individual Decision-Making and Profiling (the “Guidelines”). The Guidelines were adopted by the Working Party on October 3, 2017, for public consultation.

In its comments to the Guidelines, CIPL recommends several changes or clarifications the Working Party should incorporate in its final guidelines, including the following:

Automated Decision-Making ("ADM")

• Article 22 of the EU General Data Protection Regulation ("GDPR") should not be interpreted as a general prohibition of ADM. This interpretation is not mandated by the text of the GDPR, is not necessary to give effect to the additional protections of Article 22, and is too restrictive to ensure that the GDPR remains principle-based and future-proof in light of evolving data processing, machine learning and AI.
• Article 22 should, instead, be interpreted as a right to be invoked, which is line with the text of the GDPR as well as with the spirit and essence of the right not to be subject to solely ADM, which produces legal effects or similarly significant effects. Under this approach, ADM is permitted under all six processing grounds, including legitimate interest, unless the individual invokes his or her right not to be subject to ADM. This right can be invoked prospectively or retrospectively.
• The meaning of “legal” effect and “similarly significant” effect must be interpreted strictly to ensure Article 22 only covers truly impactful ADM.
• For an ADM process to produce a similarly significant effect, it must rise to the same level as producing a legal effect, which is a high bar.
• The Working Party should provide examples of categories of solely automated decisions that produce legal effects, similarly significant effects and decisions which do not produce such effects.
• The Guidelines on targeted advertising should be revised, and, in particular, the example put forward to demonstrate a targeted ad producing a similarly significant effect removed.
• The Working Party should clarify that using data to train or enhance algorithms is exempt from the requirements of Article 22 of the GDPR once the organization has made efforts to sufficiently de-identify the training data where possible, and employ other risk minimizing controls where appropriate and necessary (e.g., administrative and contractual controls).
• The Guidelines should clarify that the nature and scope of human intervention is highly contextual and can include a range of measures. The ultimate goal of human intervention should be to ensure correct automated processing and a fair and accurate decision, and human intervention must relate to a specific instance of ADM.

Profiling

• The Working Party should make clear that data processing only falls within the definition of profiling under the GDPR if collected personal data is actually used to evaluate, analyze or predict personal aspects relating to a natural person, and that manual evaluation does not equate to profiling.
• The Guidelines should highlight that there are two, rather than three, types of profiling:(1) general profiling and (2) profiling that results in a solely automated decision which produces legal effects or similarly significantly affects an individual.
• The Guidelines should make clear that profiling does not undermine an individual’s freedom to choose certain products or services, with the exception of extreme circumstances. Most forms of profiling and targeted advertising based on profiles do not restrict an individual’s freedom to choose certain products or services.
• Linked to profiling and targeted advertising, the Guidelines should highlight that the granularity of the segmentation does not necessarily mean the legitimate interest of the data controller is overridden automatically by the data subject. This is one factor to consider in the balancing test and must be measured alongside the risks of the likelihood of identification of an individual and safeguards provided by the data controller, such as setting a minimum audience size.
• The Guidelines should clarify that that the right to object in Article 21(1) of the GDPR does not impose a different legitimate interest standard than Article 6(1)(f), and that in the face of an objection to profiling, the data controller must only demonstrate that the data controller’s earlier legitimate interest analysis was correct and that its legitimate interests still prevail.

CIPL’s comments were developed based on input by the private sector participants in CIPL’s ongoing GDPR Implementation Project, which includes more than 85 individual private sector organizations. As part of this initiative, CIPL will continue to provide formal input about other GDPR topics the Working Party prioritizes.