Technological risk associated with the introduction of Artificial Intelligence in administrative activity can be discussed in the same terms as it is discussed at the level of private activity – issues related to, e.g., transports and autonomous vehicles, robotization, cybersecurity, and the (un)protection of personal data are similar in private and public activity.

Yet, the relation between Artificial Intelligence and administrative legal activity raises a different kind of concerns. The principle of legality and judicial control of administrative activity, as well as the safeguard of fundamental rights and guarantees of citizens, are essential pillars of the rule of Law which can be endangered by this type of technology, making it necessary to ensure the respect of these principles, rights and guarantees.

Despite its youth, the Portuguese Code of Administrative Procedure does, however, not contain many rules on this matter. Still, it establishes the principle of digital as a rule and includes the possibility of delivering documents digitally, ensuring communication between Public Administration and the procedural stakeholders (whether private or public) to be carried out remotely, facilitating the contact between the procedural subjects and contributing to a relief of public burdens in private activity. Still, excluding the aspects of facilitating the contact of citizens with public institutions and increasing transparency – cf., e.g., Articles 14, 61 and 62 of the Code of Administrative Procedure (CPA) – the mere dematerialization of processes will not lead to a significant efficiency leap if the entire administrative procedure continues to be developed in the same old way.

The real technological revolution is in the use of algorithms for electronic automation of the Administration’s decision-making processes. Such automation is expressly referred in Article 62/4 of the CPA, which generically recognizes the power of electronic one-stop shops to 'proceed with the automated issuance of acts', yet without establishing anything as to the regime for such issuance. It is precisely on this issue that the greatest challenges will arise, in particular in what concerns procedures and procedural guarantees of citizens. For this debate, it is essential to distinguish the degree of technological intervention and the type of administrative act. That will be the purpose of this intervention.

In the context of the requirement to speed up the decision-making process of administrative authorities, the use of artificial intelligence (or decision-making algorithms) is offered as one of the legitimate options, given that in many European countries the laws already allow the use of artificial intelligence in the decision-making process.

It is worth noting that the very issue of using artificial intelligence and replacing human decision-making in public administration with decision-making by means of artificial intelligence is a highly controversial issue that raises a number of problems and unanswered questions.

To mention just a few, the definition of the actual agenda on which AI can make decisions is problematic, or the question of whether AI can be used for decision-making using administrative discretion (and thus replacing human discretion with AI discretion).

The primary focus of this paper is to address issues of potential review of automated decision-making, not only by superior administrative authorities (in appeal proceedings) but also by administrative courts in judicial review.

Our basic questions are, first, how (i.e., from what perspective and with what frame of reference) such a review should be conducted, who will conduct it (a human or an appellate/judicial algorithm), and whether the use of an algorithm in the first instance proceedings will actually lead to a speeding up or slowing down of the subsequent decision-making stages (the review).

The aim of the paper is to address at the same time whether only the final product of automated decision-making should be reviewed, or whether it is even possible or realistic to review the decision-making process itself. In this context, we then ask whether it is possible (or even desirable) to subject the algorithm itself to review (in the context of the publication of the courts' decision-making practice and the possible publication of the source code of the decision-making algorithm).

In the paper, we will also address the question of how deep a review of the algorithm is desirable and whether too deep review may subsequently negatively affect the use of the algorithm itself, especially if discovering details of its settings could lead to public administration addressees submitting their submissions to the algorithm in such a form that they would be able to "trick" the algorithm.

Following our findings, we will formulate recommendations for legislative and decision-making practice and, in particular, for an effective and up-to-date setting of legislation in order to achieve and maintain a reasonably fast and at the same time sufficiently effective decision-making practice of administrative authorities and administrative courts.

The increasing digitisation of administrative procedures and reviews promises to provide citizens with faster and more potent services. From requesting a simple hearing with the county representative to participating in a public procurement process, digital solutions seem to have taken over the standard paper bureaucracy.

The transition to digital solutions in the citizen-administration relationship raises significant challenges, requiring a careful examination of both the traditional administrative principles, such as the right to human recourse, transparency and contestability, and the emerging digital rights, such as those derived from the GDPR and the recently enacted Artificial Intelligence Act.

An essential aspect of creating just digital solutions for administrative procedures is the need for increased accessibility for all citizens, irrespective of their social background, digital literacy and any physical impairments, which provides personalised, non-discriminatory solutions in the absence of a human review.

Current comparative European digital solutions for administrative procedures, their effectiveness and the need for national and European digital safeguards will also be discussed.