Facial recognition is very diffused in the East of the world but there is a particular technique strictly linked to facial recognition, that, on the opposite, is for some the product of capitalism, it is emotional recognition. It occurs when computer evaluates emotional responses, draws conclusions, and often autonomously makes decisions from a mutation in our tone of voice or from a minute movement in our face or eyes. Given the relevance of this topic, in this article, I analyse the recent draft of “Regulation on a European approach for artificial intelligence” of the European Commission, published on April 2021. I argue that this document is of revolutionary importance since numerous and significant are the innovations introduced, mainly in the video surveillance and facial recognition sectors.
Facial recognition techniques
Emotional recognition generally does not distinguish cultural and individual differences and thus can diffuse and feed stereotypes and social homologations. Because of the intrinsic danger of the indiscriminate diffusion of techniques such as facial recognition in public places, some countries and institutions like European Union have limited their massive use. The same approach should be adopted for emotional Artificial Intelligence (AI) and a series of precautions should be enacted.
In the meantime it is being regulated, emotional recognition should be forbidden in public spaces, the consent of the individual should be asked, he should be previously informed when emotional recognition will operate, but first of all the individual should enjoy the freedom to opt-out from it. To sum up, emotional AI should be used in a reliable and trustful manner and should assits human judgment and never replace it. Thus, it still needs regulation before its widespreadness.
Before the press of the new technologies, some have presaged that reckless use of AI by totalitarian regimes can create a future where a possible global totalitarian government will achieve the technological monopoly and will monitor and conditionate humanity with no limits.
Even if the hypothesis of a world-dominating power is still far and uncertain, a more probable hypothesis is that few superpowers will be able to dominate not only their own countries but the entire world through technology, having achieved the leadership in AI. If they do an unscrupulous use of AI, for example through mass surveillance or autonomous weapons, they could produce an overturning of human values that will overwhelm human rights and will conduct into human submission and maybe into human destruction. This can happen if privacy, freedom of expression, freedom of thought, free will, and freedom of movement will be conditioned and restricted.
This is a dystopian expected perspective coming from the fear that an increasing diffusion of surveilled cities will be put in place. Domestic mass surveillance is the starting point to arrive to global control and handling through algorithmic political conditioning, social media subliminal messages, and legal impositions. This could be the end of democracy. Some regions, such as the EU, have limited the use of certain pervasive technologies such as facial recognition and sanction intrusive uses of AI in our privacy or mass data collecting because they can represent a threat to human rights and democratic values. However, because human rights and democracy are not societal values in any political system, this can undermine the enjoyment of the same rights at a global level.
Governments must protect with appropriate laws citizens from the dangers that AI poses to our rights, also by ratifying international binding treaties, that establish sanctions for the transgressor States, often involved in surveillance policies indiscriminately adopted both at public and private levels.
The new EU Regulation proposal
At a supranational level, it is important to consider the recent draft of “Regulation on a European approach for artificial intelligence” of the European Commission, published on April 2021. This document is of revolutionary importance. Set as an evolution of the General Data Protection Regulation (GDPR), it addresses and solves the issues of management of AI and facial recognition systems, fixing, for its violations, penalties much higher than the GDPR, up to 6% of the global budget, against 4% of the GDPR. While there are points in common with the GDPR, numerous and significant are the innovations introduced, mainly in the video surveillance and facial recognition sectors.
With this regulatory document, Europe has intended to undertake a different path from the rest of the world in the field of legislation of AI, a new initiative but similar to that, that rendered itself a pioneer in the field of privacy and data protection: this path will be uniform to all 27 EU member States. The EU, thus, decides to regulate this field, fixing the juridical basis necessary to ensure that the AI, in the European space and market, has certain rules and specific guidelines.
Thus, outlining the European system in the field of AI, this regulation establishes the minimum standards of safety and quality in the data processing processes by AI systems, aiming to make Europe a safe environment and prone to innovation for the development of AI, but conformingly to the fundamental rights of the European citizen. Data protection regulations have been in place around the world, albeit in different forms, for a while now, but the emergence of newer technologies like AI have necessitated the need for more stringent and newer regulations to ensure fair and non-violatory play in the game of tech.
The regulation, while not prohibiting mass surveillance, introduces relevant innovations, such as the prohibition of certain practices. Firstly, the immission on the market, putting into service and use of AI systems that use subliminal techniques that go beyond the conscience of the person with consequent damage for that. Secondly, the immission on the market systems that exploit the vulnerability of a specific group of people because of their age, mental or physical disability in order to materially falsify the behavior of a member of the group with consequent damages. Then, the immission on the market of AI systems by or on behalf of public authorities for the assessment or reliability of physical persons on the basis of their social behaviour or of their personal characteristics, with the assignment of social scoring, that produces certain results, such as the detrimental treatment in social contexts unrelated to those in which the data were generated or the unjustified damaging treatment or disproportionate in regard to the person’s social behaviour or to its seriousness. Finally, the use of automated “real-time” remote biometric identification systems in accessible public spaces with the purpose of law enforcement, except for public security finalities and subject to reasoned autorization given by an independent judicial or administrative authority.
Within the guarantees introduced by the draft of the regulation, there is also the guarantee concerning certain systems, such as chatbot, which will be subjected to minimum transparency obligations, aimed at allowing those who interact with the content to make informed decisions. To this end, the user will always be made aware of the use of an AI system and will therefore be able to freely and consciously decide whether to continue or to withdraw from using the application. On the opposite, the new rules will not be enacted to the systems falling into the category of the “little or no risk”.
However, as said, the new regulation presents also significant points of contact with the GDPR, such as the prevision of the guarantee of accuracy, robustness, resilience, and security of AI systems, just as the GDPR guarantees the confidentiality, precision, and safety of personal data, see for example article 32.
Concluding remarks
The proposal of European Regulation of AI offers therefore to guarantee a continuity path along with the GDPR, being an evolution aiming to better face the risks that have occurred in the technology sector due to the pandemic. This European legal proposal will not be able to have a deep impact even on a global level, as it was with the GDPR, ensuring the AI systems used in the EU will be safe, transparent, ethical, impartial and under human control and indeed, anything considered a real threat to the European citizen will be banned.
The European strategy pursues four goals: to establish the enabling conditions in favour of the development and diffusion of AI, to build a strategic leadership in high impact sectors, to do of EU an ideal environment for AI to flourish, and to make sure that AI technologies are to serve the individual. The European recovery after the pandemic cannot avoid a technological evolution that is attentive to the environment, to the society, and to the individual.
In the name of cultural relativism, we cannot accept the violation of common values. It is crucial that international pressure can persuade also certain totalitarian regimes through a multilateral international convention on AI rather than with a universal declaration because only an international treaty is legally binding. Moreover, this is an exquisitely utopian forecast difficult to realize in the present geopolitical scenario. The international community should go in this direction.
Maria Stefania Cataleta is an Italian lawyer, researcher, and lecturer in international law, international criminal and human rights law in several universities, including La Sapienza University, Roma Tre University, and Niccolò Cusano University in Italy, and LADIE-Université Côte d’Azur in France. She holds Ph.D. in Public Law from Université Côte d’Azur (France), Ph.D. in Political Science from Roma Tre University (Italy), and LL.M. in International Law from City University of London (UK).