Artificial intelligence: country needs to adapt legal framework – 03/08/2023 – Tech

City Hall’s intention São Paulo’s decision to use facial recognition cameras to monitor infractions exposes a serious regulatory gap: Brazil has no rules for the use of artificial intelligence (AI).

Solutions based on artificial intelligence already cause disruptions in the labor market, intellectual property and privacy, and there are more than 700 startups specializing in artificial intelligence in the country — this is the most recent data from the Distrito platform, from 2021, but the number must have grown with the worldwide phenomenon ChatGPT.

For specialists, it is necessary to protect consumers, employees and citizens and, at the same time, guarantee legal security for innovative businesses.

Brazil is still in the early stages of developing its strategy and regulatory framework for AI, when compared to developed countries and China.

In the Senate, three legislative proposals are being processed to close this legal vacuum. They await a dispatch from the Bureau’s Secretary General, Gustavo Sabóia Vieira, who responds to the president of the House Rodrigo Pacheco (PSD-MG). The next destinations for these texts will be the competent committees, where the rapporteur for the agenda will be appointed.

Bill 21/2020, already approved in the Chamber of Deputies, sets out principles that should guide the application of AI in the country — respect for human dignity, transparency in algorithms and protection of personal data. Now, it is under consideration by senators, under number 872/2021. PL 5.051/2019 has a similar text and was added to the agenda.

The approval of PL 21/2020, under criticism from specialists for the speed with which it advanced and for its vagueness, led the Senate to establish, in February 2022, a commission of jurists to debate the regulation of artificial intelligence. This group of specialists prepared a substitute draft to complement the project that arrived from the Chamber.

This replacement draft prepared by 18 jurists takes a more normative approach and recommends banning facial recognition technologies and social ratings — a note similar to a credit score, aimed at behavior in public.

The 900-page report prepared by the commission of 18 jurists was approved last December and will serve to support the discussion of the ongoing proposals. The future project rapporteur will decide whether to present the substitute.

The Brazilian text innovates by proposing a system of individual rights, according to InternetLab director Mariana Valente, who participated in the group of specialists. It would be possible, for example, to obtain compensation from the AI ​​platform administrator in case of damage caused by the system.

For Abria (Brazilian Association of Artificial Intelligence), too rigid regulation can harm companies’ innovation and competitiveness. “It is necessary to strike a balance between the need for regulation and the promotion of innovation and competitiveness”.

A member of the committee of jurists, the founding director of Data Privacy Brasil Bruno Bioni says that, according to the suggestion sent to Congress, the rigor of the rules would vary according to the risk linked to the artificial intelligence resource under discussion. Recognizing criminals would be too risky and should be banned. For employee recruitment, the risk would be high and transparency would need to be reinforced with an explanation of how the algorithm works. Financial scores for extending credit would have medium risk.

Why is regulation important?

The regulatory apparatus is largely responsible for bringing legal certainty to an economic activity.

According to Clara Keller, entrepreneurs need predictability in relation to business risks, and the organization of the sector allows for greater planning and public and private investment. “A safe and comprehensive regulatory framework would be an element in favor of the development of these technologies in Brazil”.

For Bioni, from DataPrivacy Brasil, it is a fallacy that regulation can hinder innovation. “It can be a driving force behind innovation when standards are developed for that purpose.” The LGPD, for example, guaranteed security in business for data operators.

FGV (Getúlio Vargas Foundation) professor Luca Belli says, however, that Brazilian legislators should consider the local economy before approving stricter regulation. “This configuration for me is the worst: copying Europe’s regulation and not producing the same innovation as China or the United States”.

Does any country already have a regulatory framework for AI?

There is no country with a regulatory framework focused on artificial intelligence. There are proposals at different stages of discussion as in the European Union and Brazil and state laws regarding specific AI features as in the United States.

In the Netherlands, for example, the government bequeathed the responsibility for dealing with violations related to artificial intelligence to the ANPD (National Data Protection Agency) body.

In China, regulation is up to the Cyberspace Administration of China, responsible for the internet in the country.

What does the European Union proposal say?

In the proposal under analysis in the European Union, artificial intelligence is seen as a product, which needs to undergo evaluation and certification processes before entering the market.

Uses of AI that represent risks considered unacceptable are banned: systems with the aim of manipulating people, exploiting the vulnerability of specific groups, social scoring features and biometric identification in real time and at a distance.

Researcher at the Leibniz Institute for Media Research Clara Keller says the current European regulatory project is criticized for treating AI as a mere product, ignoring how these technologies, once integrated into our societies, can affect rights and legal relationships in multiple ways and not always desirable.

The European model does not present a system of individual rights, like the Brazilian proposal.

Local businessmen have been pressing against the passage that would determine the mandatory understanding of artificial intelligence algorithms by the population. This group argues that the condition would make business unfeasible, because AI works through probabilistic operations intelligible only to machines, according to Luca Belli, from FGV.

How is the regulation in the United States?

North American legislators have left the responsibility for creating specific rules for artificial intelligence resources to the states.

According to Bioni, from Data Privacy Brasil, this regulation also varies by sector and use. The city of San Francisco, California, for example, prohibits surveillance systems with artificial intelligence and identification of biometric data.

The companies themselves also develop internal governance to mitigate risks in technology development. Google says it follows 10 principles in developing artificial intelligence applications, including avoiding reinforcing bias, promoting safety testing and public understanding.

How is regulation in China?

In China, the internet regulatory body forces technology companies to reveal their algorithms, including artificial intelligence. Since March 2022, the Cyberspace Administration of China has been reviewing whether the resources used by local platforms are moral, ethical, transparent and accountable.

The Chinese government, on the other hand, creates opportunities for small companies to test innovative technologies without having to fully comply with current regulations —the model is called regulatory sandbox and aims to encourage innovation, according to Luca Belli, from FGV.

What uses of AI are already being made in the country without prior regulation?

In Brazil, artificial intelligence technologies are already employed to calculate credit scores at financial institutions, screen candidate resumes and in social media algorithms such as TikTok.

Local governments also contract for these services. The state of São Paulo uses facial recognition technology to monitor Metro line 3. The São Paulo City Hall has been trying to implement a monitoring system with 20,000 cameras embedded with artificial intelligence devices, which could violate minority rights, according to the Municipal Court of Auditors.

The Secretariat of Public Security of the Federal District also uses artificial intelligence to increase the accuracy of electronic anklets and issue alerts in case of abnormal behavior by detainees.

Facial recognition, some content moderation tools used in social networks, and in some cases these systems are even used within the scope of public services.

While a legal framework is not approved, who regulates AI in Brazil?

State bodies need to enforce current legislation. The ANPD (National Data Protection Authority), for example, needs to monitor the processing of data used to train the neural networks behind the new algorithms.

According to the director of the board of directors of the ANPD (National Data Protection Authority), Miriam Wimmer, the ANPD can choose to regulate practices in artificial intelligence or just issue guides, recommendations and practical guidelines to facilitate compliance with the LGPD (General Data Protection Law). Data) by organizations. This is not yet defined.

Cade (Administrative Council for Economic Defense), Senacon (National Consumer Secretariat), Anatel (National Telecommunications Agency) and Labor Justice have to do the same in their respective areas.