*Per Anderson Rohe
The European Union (EU) approved on June 14 of this year a pioneering project on the regulation of Artificial Intelligence (AI)[1], since they will be the first rules in the western world about the theme[2]. Since, soon, China will have its own rules, competing in parallel, motivated by the desire to regulate Generative AI (like chatGPT, unavailable in the country), undergoing a safety inspection before marketing it in Chinese territory[3].
The EU itself accelerated its AI regulation process, which started in 2021[4], motivated by the “moral panic” generated around the advancement and rapid popularization of Artificial Intelligence (chatGPT, above all)[5]. First with predictive AI (facial recognition and public surveillance). And now, generative AI, producer of texts and synthetic images.
Even so, the plans for the European regulation to come into force are only scheduled for 2026[6]. For this reason, many experts defend the idea of a provisional and voluntary code of conduct, given that the high risks of AI could not wait for laws that are not yet in force.[7].
The standard approved by the European Parliament, after 499 votes in favor, 28 against and 93 abstentions[8], which will regulate the use of AI in the European bloc, is now, however, entering a complicated and decisive phase, as it needs to be negotiated with all representatives of its twenty-seven member states. A long and complex process, to be followed closely, as not all countries are at the same level of leveling and/or institutional maturity in terms of data protection, privacy and the environment, regulatory authority autonomy, information security and digital sovereignty. And, above all, which safeguards will be prioritized in the specific case: whether of legality, legitimate interest of the State, necessity, reasonableness, proportionality, among others.
THE classification by risk of each application of Artificial Intelligence proposed by the EU, and already suggested at the time of the public consultation by the Senate, about the regulatory framework for AI in Brazil[9] – is well received by some sectors[10], although there is still no consensus on the effectiveness of this classification criterion or categorization into low, medium and high risk.
In the case of high risk, for example, the danger to which one is currently exposed makes it unacceptable or more difficult to be tolerated by civil society, especially in an environment without safeguards, specific rules or where there is a legislative gap regarding what, in theory, it would threaten democracy, human rights, individual freedom and fundamental guarantees. As in the case of AI emotion recognition systems, predictive policing and biometric surveillance in public places that threaten citizen privacy and data protection[11].
And that, therefore, the framework European should, or at least should, be received with more caution (see the precautionary principle), since the risk approach (known as risk-based approach)[12] will vary second: a) the sector involved, such as health – which necessarily involves confidential and sensitive personal data – or b) the area to be applied, ranging from video games to autonomous cars[13].
Not by chance, AI specialists are now divided between considering the European regulatory model a promising model and one that will legitimize similar regulatory processes in other countries[14]. Mainly due to the influence of the so-called “Brussels Effect”, when taking European regulation as the best example of good international practices in the sector (benchmarking). Already others[15] they prefer to wait for its developments, since there cannot be an automatic replication or mirroring process, along the lines of “copy and paste”, since local specificities need to be considered before such a model becomes, in fact, public policy domestically.
Brazil, unlike the European Union, and as an example of a developing country, traditionally more a consumer than a producer of technological innovation (and therefore more vulnerable and susceptible to risk), affected by the fragility of institutions, urban violence and insecurity public, cannot afford to stimulate (via nugde or incentive government policies) the exclusive use of the internet by mobile devices, approach for payments, migration and storage of personal data in the cloud, since the theft of cell phones is a national reality in large cities such as Rio and São Paulo . It is enough, therefore, to "crack" the password or take a smartphone unlocked to gain access to an entire life history and consumption habits of the individual victim of crime or infraction.
The problem, then, to be faced ahead will be in the criteria of each body, entity or country to assess this risk between tolerable (generally of low and medium impact) and intolerable (high and/or unacceptable risk). And the stage of use and development of the new technology. Which will certainly vary from the practical and everyday reality of one region to another. Whether North or Global South country. And, therefore, depending on the existing mechanisms for measuring impact and risk classification, auditing and algorithmic transparency for its effective implementation.
*Anderson Röhe is a fellowship researcher at the Artificial Intelligence WG at Think Tank ABES
References:
[1] European Commission. Proposal for a REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
LAYING DOWN HARMONIZED RULES ON ARTIFICIAL INTELLIGENCE (ARTIFICIAL INTELLIGENCE ACT) AND AMENDING CERTAIN UNION LEGISLATIVE ACTS. available in this link.
[2] European Parliament. AI Act: a step closer to the first rules on Artificial Intelligence. available in this link.
[3] AFP apud G1 Globo. China wants to regulate the use of artificial intelligence. available in this link.
[4] AFP apud Istoé. European Parliament approves project to regulate the use of AI. available in this link.
[5] IT INSIDE Online. Why Big Data without curation is worth little with the advancement of artificial intelligence. available in this link.
[6] Op. cit. available in this link.
[7] JN Agencies. US and EU announce joint code of conduct for Artificial Intelligence. available in this link.
[8] Op. cit. Available at this link.
[9] GEDI Mack – LAW AND INNOVATION RESEARCH GROUP. Contribution to the Public Consultation on the Regulatory Framework for Artificial Intelligence. available in this link.
[10] CNN Brasil. Artificial intelligence regulation is urgent and complex, says expert. available in this link.
[11] Op. cit. available in this link.
[12] Op. cit. available in this link.
[13] Op. cit. available in this link.
[14] Op cit. available in this link.
[15] Rohe, Anderson. The Future in Prospect before a Global Declaration for the Internet. available in this link.
EN 
PT 
