IBM: Any AI regulation must allow innovation

US tech giant IBM argues that any regulation of artificial intelligence in Brazil must be well thought out so innovation is not hindered, and business AI applications are separated from individual AI usage.

“In the context of generative AI, we are obsessed with ethics and security. We try to take this vision to the authorities. And we seek to make this separation of what is AI for the world in general and what is AI for business,” Fábio Mucci, security leader at IBM Brasil, said in reply to BNamericas during a press conference in on Thursday.

IBM is directly and indirectly involved in discussions with Brazilian authorities regarding AI.

Indirectly, the company has worked with Brasscom, the Brazilian association of IT companies, and, directly, has interacted with the government through its regulatory affairs department in Brasília, in different contexts.

“I’m very close to the AI discussion in Brasília, with the ministries. There is a concern regarding the limit of regulating the topic,” said Mucci. “There is a fine line between over-regulating and letting technology grow and evolve.”

Brazil is currently discussing a proposal to regulate AI presented by senate president Rodrigo Pacheco last year.

At the end of 2023, sessions and hearings of the working group created by the senate to discuss the proposal were extended for a few more months. The postponement was in response to requests from technology companies, which wanted more time to discuss the matter.

Jorge Sukarie, board member at Brazil software association Abes, told BNamericas this week that the regulation cannot be rushed. He compared the topic to the several years of discussion that led to the approval, in 2014, of Brazil’s internet framework (Marco Civil da Internet).

At the end of last year, President Luiz Inácio Lula da Silva issued a decree creating a new national cybersecurity policy, as well as the national cybersecurity committee (CNCiber). The latter will be chaired by the institutional security office of the presidency.

Last week, the government launched calls for the composition of the rest of the committee, to include members from civil society; scientific, technological and innovation institutions; and companies, with one chair each.

“One of the topics that’s been actively discussed in these early days of CNCiber is precisely generative AI, in the context of security. And I think we will soon have some news on this regard,” Mucci told reporters.

From a broader security perspective, Mucci said genAI should not bring mean changes in cybercrime as it is today. However, it will scale it up. 

Among the main threats cited, some of which are already known, those expected to grow include phishing scams powered by AI, quantum computing accelerated decoding, and malicious LLM (large language models), or ‘FraudGPT’.

FraudGPT is being described as tools for creating undetectable malware in the writing of malicious code.

PUBLIC SERVICES

Beyond regulation, Mucci defends advances in the use of AI to streamline services provided to citizens.

This week, Dataprev, the Brazilian social security and pensions data processing company, launched the Atestmed platform. The system uses artificial intelligence algorithms to cross data and scan documents sent digitally by beneficiaries applying for the seguro-doença health coverage.

In addition to streamlining the process, Atestmed seeks to reduce fraud in the documents, a longstanding problem in the social security system.

Mucci sees increasing synergies between data from public companies Serpro, INSS, IRS and Dataprev. Yet, he said many other public federal government databases still lack integration.

“We look at the context of the public machine, at the immensity of available data, and this is a path of no return,” he said about Dataprev's initiative.

“And I think the use of generative AI will help a lot.”