Pending regulation, Brazil starts to revise national AI strategy

Brazil is reviewing its national artificial intelligence strategy, which was launched in 2021, while a proposal for specific regulation of the technology remains pending in congress.

The review process is led by the science, technology and innovations ministry (MCTI). The AI strategy aims to guide the country's priorities around the topic.

“With the initiative, MCTI will revisit objectives, axes, goals and actions to align it with national interests and priorities. The EBIA [strategy] review process should be completed by May 2024,” the ministry said in a statement.

Since the strategy's launch, MCTI has supported the installation of 10 applied artificial intelligence research centers in the country. These centers are dedicated to developing research aimed at solving problems using AI, in addition to contributing to the training of human resources in the sector, according to the ministry,

Science and digital transformation technology secretary Henrique Miguel said the review will focus on “what Brazil wants, wishes for and expects from the impacts that artificial intelligence brings.”

The review will also touch on aspects such as production, development, use, regulation, security and cooperation between Brazil and other countries.

GOVERNING AI

The revision comes as MCTI is chairing the AI Axis, which is part of the G20 Digital Economy Working Group, as Brazil holds the presidency of the group of the world's leading economies.

EU officials just achieved an agreement on what is considered the world's first comprehensive AI law, which could influence Brazilian debates. 

Technical details and specific rules are expected to be detailed soon, but among other things, the EU's proposed legislation sets risk categories for AI systems. 

It also enforces transparency, while establishing that companies must comply with EU copyright law and share detailed summaries about content used for training.

Large language model (LLM) platforms such as OpenAI’s ChatGPT will have to conduct model evaluations, assess and mitigate systemic risks, conduct adversarial testing, report to the European Commission on incidents, ensure cybersecurity and report on their energy efficiency.

In the Brazilian congress, negotiations continue for the creation of AI guardrails.

Senate president Rodrigo Pacheco submitted an AI bill in May. The bill is a result of recommendations from legal expert members of a working group created in 2022 and centered on five AI topics: principles, the rights of those affected, risk classification, governance obligations and requirements, and oversight and accountability.

The Brazilian text is also inspired by the European proposal.

Senator Eduardo Gomes, head of the communications and digital rights committee and the bill's rapporteur, told BNamericas in September that a vote could take place this year, which is now unlikely.

Gomes expected to accelerate the 15 public hearings planned to debate the text, but the attempt was not successful.

Furthermore, big technology companies and associations created the Coalition for Innovation and Responsibility in Artificial Intelligence, and issued last month an open letter calling for a longer debate on the bill.

Fearing a hastily approved legislation that could restrict innovation, they call for further discussions so that regulation “does not become obsolete with new developments” regarding the technology.

The letter is signed by Câmara-e.net, which includes companies such as Google, Amazon, Meta and MercadoLibre; by Brazilian software association ABES and AI association ABRIA, among others.