Approval of the Artificial Intelligence Framework is criticized by experts – Legislation – Tecnoblog

On Wednesday afternoon (29), the Chamber of Deputies approved the legal framework for Artificial Intelligence, which regulates AI sector practices in Brazil. The text goes for approval by the Federal Senate. Experts heard by Techblog, however, diverge on the content of the PL that defines the principles to be followed by private and public entities that use smart technologies such as machine learning.

Artificial Intelligence: regulatory framework criticized by experts (Image: Geralt/Pixabay)

AI landmark is against government regulation

The bill voted in the House states that any system that relies on computational processes to learn from the external environment, making predictions, recommendations, classifications or even capable of taking decisions autonomously can be classified as AI.

The initial text of the framework was presented by Deputy Eduardo Bismarck (PDT-CE). The bill underwent some changes when it passed to the rapporteurship of Deputy Luiza Canziani (PTB-PR), when it was processed by the Chamber of Deputies’ Committee on Science and Technology, Communication and Informatics (CCTCI).

The proposal also points out that the federal government should not act as a regulatory agent for any Artificial Intelligence solution, acting only when “absolutely necessary”. The text also provides that the role of the Union must be the incentive: when the development of technology is of low risk, it must promote research in favor of the IA.

Finally, the PL points out that, when government regulation on AI is applicable, it must come from agencies and supervisory bodies, which act in the regulation of their respective fields, such as Anvisa (National Health Surveillance Agency) or ANAC (Agency National Civil Aviation Agency).

Experts heard by Techblog differ on the content of the framework for Artificial Intelligence.

Bill can generate investment, says lawyer

For the lawyer Ricardo Freitas Silveira, a partner at LBCA, and a doctoral candidate in Constitutional Law at the IDP, the bill approved yesterday afternoon places Brazil at an important level compared to other OECD (Organization for Cooperation and Development) countries. that the framework that regulates AI is a guideline of the international body.

In an interview with Techblog, Ricardo Silveira says:

“The Artificial Intelligence milestone is a first step, because once again we are following the OECD guidelines, this brings legal certainty and allows investment, especially following the European standard.”

Silveira also points out that what he considers a “black box” should not be created around AI: little transparency in systems that are part of technologies implemented on a daily basis, such as financial systems, healthcare systems or even entertainment.

“It’s an important step forward, because it will require companies to comply with the principles when the law actually comes into force. This benefits the market in general”, completes the lawyer.

Proposal did not go through public consultation

President of the Chamber of Deputies, Arthur Lira (PP-AL) during a session on Wednesday (Image: Michel Jesus/ Chamber of Deputies)

On the other hand, Bruno Bioni, director and founder of the Data Privacy Brasil institute, points out that the text of the Chamber’s Artificial Intelligence framework is insufficient. The expert’s criticism is as follows:

“All the tools and mechanisms for public participation have not been exhausted. The discussion around the PL compared to others that legislated around innovative themes — such as the Marco Civil da Internet and the LGPD (General Data Protection Law) — was smaller, there were fewer public hearings and the text was not placed under public consultation on the Chamber’s website at least once.”

Also according to Bioni, the text lacks a high degree of accountability. This means that the PL does not allocate rights of citizens and duties of the Public Power and private companies that develop solutions in machine learning, for example. Furthermore, the law does not establish the due responsibility of those who control and offer the AI ​​to the market.

One of the most controversial points in the approval of the AI ​​framework is Art 6. This section of the text has an item (6) that implies that civil liability for damages caused by the AI ​​for any damage is subjective and not objective. This means that the victim must prove that the action of an operator of any such technology was at fault. It would not be enough, for example, just to prove that there was damage.

For Bioni, this article is inconsistent with the rest of the project:

“If the PL wants to create a safe and beneficial system for AI, this stretch is inconsistent and incongruous. The law wants to avoid compensation and punishment for any errors that cause harm to victims, defending who has more economic power.”

PL can generate legal uncertainty

Attorney Christian Perrone, from ITS Rio (Institute of Technology and Society of Rio de Janeiro), agrees with the criticism that there was much less time to discuss a bill that “want to achieve a lot”. In Perrone’s assessment, the law does not have the necessary depth to debate such an innovative topic:

“At the moment, it would seem positive that we have a milestone [de IA] fast, but there are downsides: the discussion on the subject did not have time to properly clarify the concepts present in the PL, which define such an innovative sector and vast possibilities. AI can be used both in a simpler chatbot and to define who is a priority in the organ donor queue, or what the surgery will be recommended for the patient. The PL, as it is, can provoke future discussions, which in turn generates legal uncertainty.”

Big techs reportedly lobbied for quick approval

A source who declined to be identified cites that the speed and prevailing voters in favor of the project — the AI ​​milestone was approved by a score of 413 to 15 — is due to an old practice not so innovative: lobbying. Apparently, big techs such as Google and Facebook have streamlined the proposal’s processing within the Chamber of Deputies and lessened resistance to approval.

“Who can be held responsible if an AI system causes damage: who develops, who sells and who buys,” said the expert heard by Techblog.

The PL of the Artificial Intelligence framework had the “blessing” of the Bolsonaro government to pass through the Chamber. This was signaled by the director of Science, Technology and Digital Innovation at MCTI (Ministry of Science, Technology and Innovation), José Gontijo. He praised the project at an event promoted by Digital Convergence:

“The substitute for Deputy Luisa Canziani (PTB/PR) was a very good text, serene, principled. It gives some guidelines and scope for the government to regulate some things, but it doesn’t immediately generate an impact, preventing innovation.”

Leave a Comment