The Brazilian dilemma of AI regulation – 04/09/2024 – Tech

It is no exaggeration to say that there is a global race to regulate artificial intelligence. In March, the European Parliament approved its AI Regulation. The United States and China have also discussed related legislation.

In Brazil, two legislative proposals stand out for regulating the matter: one influenced by European standards and the other focused on ethical issues, raising the debate on whether or not to adopt the European model.

AI regulation differs from other areas for one reason: there is no significant experience on the topic anywhere in the world. While in the field of data protection Europe has had legislative maturity of almost 50 years, the regulation of AI has been a path followed blindly. It is not yet possible to know whether the model adopted will have a positive or negative impact on the bloc, especially with regard to innovation.

In the case of data protection, Brazil was right to be inspired by the already mature European model. But when it comes to AI, which direction should we go? First of all, caution is needed.

The first step is not to relegate its own regulation to the market. The second, and most important, in creating regulatory designs suitable for the current moment of uncertainty.

The European model is uncertain, as modern regulatory law is only effectively implemented in legislated form after the accumulation of experience on risks and damages. Air traffic was only regulated after planes began to fly, and empirical evidence has shown where damage arises or can arise; regulating at the previous time would affect the development of the technology and its benefits.

The same occurs with AI: despite the urgency of the issue, concrete damages are still incipient and sectoral, and can be absorbed by existing regulatory structures.

For example, the controversial advertisement with a deepfake by singer Elis Regina should not imply the need to create a new law, as Conar’s regulatory apparatus is capable of dealing with the issue.

Following the AI ​​Act model implies, given the little regulatory experience, carrying out a dangerous exercise in futurology on one of the most important issues of our era. An ideal regulatory regime should sectorize regulation, creating mechanisms so that different sectors can monitor and inspect new technologies, issuing alerts if they cause damage or risks.

It is not reasonable to believe that a general law would effectively regulate different sectors, such as health and transport, agribusiness, education, etc. without stifling innovation. Sectorizing AI regulation means giving regulatory agents in each sector the possibility of building the most appropriate regulatory apparatus for their reality.

The creation of a regulatory authority for AI, another legislative innovation, is also foolhardy. The most rational thing would be for sectoral regulatory bodies to look into the issue, with the creation of specialized committees within regulatory agencies, or even a body that acts as a sector coordinator, but not as a centralized regulator.

There is a huge gap between the media perception of the dangers of AI, the concrete harms of the new technology and the regulatory experience available on the topic.

The Brazilian dilemma lies precisely in the fact that our regulatory option means, above all, a decision regarding our geopolitical positioning in the global market for AI innovations.

With the G20 taking place in Brazil, we have a global showcase to display signs of maturity, rationality and pioneering in technological regulation, aligning innovation and protection of rights in a scenario of uncertainty.