{"id":99214,"date":"2026-01-14T08:00:35","date_gmt":"2026-01-14T11:00:35","guid":{"rendered":"https:\/\/abes.org.br\/?p=99214"},"modified":"2026-02-05T19:35:36","modified_gmt":"2026-02-05T22:35:36","slug":"o-ultimo-lugar-onde-o-brasil-deveria-inovar-e-na-regulamentacao-da-inteligencia-artificial","status":"publish","type":"post","link":"https:\/\/abes.org.br\/en\/o-ultimo-lugar-onde-o-brasil-deveria-inovar-e-na-regulamentacao-da-inteligencia-artificial\/","title":{"rendered":"The last place where Brazil should innovate is in the regulation of Artificial Intelligence."},"content":{"rendered":"

\"\"* By Rodolfo F\u00fccher<\/em><\/p>\n

There are moments in history when a country's decisions define not only its economy, but its place in the future. The regulation of Artificial Intelligence is one of those moments, and perhaps the last place where Brazil should try to "innovate".<\/p>\n

In the movie\u00a0Apollo 13<\/em>, The phrase "failure is not an option" has become a symbol of overcoming risk. In our case, making a mistake in the regulatory path could mean being left out of the next technological cycle.<\/p>\n

Bill 2338\/2023, which aims to create a regulatory framework for AI in Brazil, starts with good intentions. But the current text threatens to turn prudence into paralysis, and what should protect the future may end up making innovation unfeasible.<\/p>\n

The first worrying point is the\u00a0requirement to track all copyrighted content used in training AI models<\/strong>.<\/p>\n

Requiring developers to publish a summary of all copyrighted content used in the development process on an easily accessible website is, in practice, the same as saying:\u00a0Don't do anything in Brazil.<\/em>.<\/p>\n

Modern models are trained on trillions of data points. No leading technological country, such as the United States, Japan, or Singapore, imposes a similar obligation because it is technically unfeasible.
\nTo insist on this is to condemn Brazil to technological dependence: we stop producing knowledge and start consuming what others create.<\/p>\n

The second point is the\u00a0reversal of the burden of proof<\/strong>.<\/p>\n

At first glance, it seems reasonable that the judge should have the power to reverse the burden of proof when the victim is economically disadvantaged or when the characteristics of the AI system make it excessively difficult to prove the requirements of civil liability. In practice, the developer or supplier would then have to prove that they did nothing wrong.<\/p>\n

In an already overburdened judicial system, this change would pave the way for a flood of compensation lawsuits and increased insurance and compliance costs for operating in Brazil.<\/p>\n

The result would be to discourage precisely those who invest, research, and create, the opposite of what the country needs.<\/p>\n

The third point is the\u00a0generic and horizontal regulation<\/strong>, replete with subjective definitions of risk and obligations, which increases the potential for litigation.<\/p>\n

The text stipulates that it will be up to the sectoral authorities to define, in the future, when the obligations may be relaxed or waived.<\/p>\n

In practice, an extensive list of requirements is imposed from the outset, which increases development costs and hinders projects, only for the agencies to later dictate what should never have been mandatory.<\/p>\n

Wouldn't it be more logical for these same agencies to define, from the outset, what really makes sense to regulate and what can be made more flexible?<\/p>\n

Meanwhile, the world is going in a different direction.<\/p>\n