Three years after the launch of ChatGPT, Wall Street is beginning to question the sustainability of the euphoria surrounding artificial intelligence. Meanwhile, Brazil is finalizing its AI regulatory framework and launching public career guidance platforms. This contrast between financial speculation and institution building reveals something fundamental: we are leaving behind the era of promises and entering the phase of delivering real value - with all the risks and opportunities that entails.<\/p>\n
According to a report by Bloomberg via InfoMoney<\/a>, signs of skepticism are proliferating in the American financial market. The recent fall in Nvidia<\/strong> and the negative feeling around the Oracle<\/strong> - which saw its shares plummet after reporting increasing spending on AI with no commensurate return - illustrate a central dilemma: the rubber has finally met the road.<\/p>\n Jim Morrow, CEO of Callodine Capital Management, sums it up: \u201cIt was a good story, but now we're taking a gamble to see if the return on investment will be good.\u201d<\/em><\/p>\n The numbers are impressive and frightening. A OpenAI<\/strong> plans to spend US$ 1.4 trillion<\/strong> in the coming years and expects to burn US$ 115 billion by 2029 before generating positive cash flow in 2030, according to The Information. Meanwhile, giants such as Alphabet, Microsoft, Amazon and Meta<\/strong> must collectively invest more than US$ 400 billion<\/strong> in capital over the next 12 months, mostly in data centers.<\/p>\n The problem? AI-related revenue growth at these companies is still far from covering costs. Worse: depreciation expenses for the data centers created in the AI fever have risen from US$ 10 billion to US$ 22 billion quarterly in less than a year, with a projection of reaching US$ 30 billion. Meta and Microsoft<\/strong> should have negative free cash flow in 2026 after considering returns to shareholders.<\/p>\n I've worked with companies that have invested heavily in technology without validating the real return. I've seen up close how the pressure to \u201cbe on the bandwagon\u201d can lead to risky decisions. The difference now is scale: we're not talking about startups - we're talking about the biggest corporations on the planet betting trillions with no guarantee of a return.<\/p>\n While Wall Street questions the bubble, Brazil is moving in another direction: institutionalization. According to Economic Value<\/a>, the text of the Bill No. 2.338\/2023<\/strong> The bill to create a regulatory framework for artificial intelligence is expected to go before the full House later this year.<\/p>\n The MP Aguinaldo Ribeiro<\/strong>, The bill's rapporteur confirmed that the text will incorporate the bill creating the National AI System (SIA)<\/strong> - proposed by the federal government - and also the Redata<\/strong>, a special tax regime for data centers in Brazil. The Minister of Science, Technology and Innovation, Luciana Santos<\/strong>, highlighted a critical piece of information: 60% of Brazilian data is processed outside the country<\/strong>.<\/p>\n This issue isn't just technical - it's strategic and sovereign. Processing data locally means protecting information, generating skilled jobs and creating economic value here. O Brazilian Artificial Intelligence Plan (PBIA)<\/strong> foresees an investment of R$ 23 billion by 2028<\/strong> focused on solutions that improve people's lives and innovate in public service.<\/p>\n In parallel, the SENAI has launched a free AI platform<\/a> developed with Google Cloud<\/strong> for professional guidance and employability. The tool, called Nai<\/strong>, analyzes CVs, professional history and career aspirations, cross-referencing this information with real market demands to recommend personalized training and identify job opportunities.<\/p>\n This is AI applied with a clear purpose: to reduce the skills gap, connect people to opportunities and democratize access to qualifications. It's not hype - it's social infrastructure.<\/p>\n While Brazil builds governance, the Goal<\/strong> is advancing on its users' data. According to G1<\/a>, As of December 16, user interactions with Meta's AI will be used to target ads and suggest posts on Instagram, Facebook, Threads and WhatsApp.<\/p>\n The company will also use public information from Threads to train its AI systems. Users can object to the use of their data, but few know how or are even aware of the change. O Consumer Protection Institute (Idec)<\/strong> criticized the measure for lacking transparency. The concern is legitimate: Meta is able to infer cultural, economic, social interests and even sensitive aspects from conversations with its chatbots.<\/p>\n THE National Data Protection Authority (ANPD)<\/strong> had already suspended the measure in the past, releasing it only after adjustments. Now, Meta is expanding the scope of collection once again. Trust, as I've said many times, is the scarcest asset in the digital age. And once lost, it rarely returns.<\/p>\n While some speculate and others legislate, Brazilian investment managers are already operating with AI on a daily basis. According to Estad\u00e3o E-Investidor<\/a>, houses such as Ita\u00fa Asset, BB Asset, Santander Asset and Bradesco Asset<\/strong> use AI to analyze corporate reports, capture market sentiment on social networks and even write management letters.<\/p>\n But there is one fundamental detail: human supervision remains at the heart of decision-making<\/strong>. Guido Chagas, from Santander Asset, warns: \u201cIf you let it make decisions on its own, in a matter of time the tool will make a big mistake.\u201d<\/em> Juliano Santos, from BB Asset, estimates that only 10% to 20%<\/strong> of the models tested are approved for effective use in portfolio management.<\/p>\n In my work with companies and governments, I observe that the real competitive advantage is not in having AI - it's in knowing how to use it with discretion, governance and adequate supervision. AI without human judgment is not efficiency - it's institutional roulette.<\/p>\n THE Softex<\/strong> launched its third policy brief<\/em>, entitled \u201cRegulation that Competes: Reliable AI to Scale Brazil's Productivity\u201d<\/a>. The document proposes that regulation should act as a vector of productivity and confidence, adopting a model based on risk levels.<\/p>\n Among the priority recommendations for the next two years are the creation of the National AI Evaluation System<\/strong>, the launch of the IA Trusted Seal - Brazil<\/strong> and a competitiveness package for micro, small and medium-sized enterprises (MSMEs), including compliance vouchers and regulatory implementation kits.<\/p>\n Trust is key to scaling AI adoption, especially in sensitive sectors such as Healthcare and Digital Government, which require traceability, human oversight and transparency. Regulating AI isn't a brake on progress - it's protection against barbarism.<\/p>\n A worrying phenomenon is beginning to emerge in the Brazilian judiciary: the Shadow AI<\/strong>. According to CONJUR<\/a>, This term describes the use of AI tools without the knowledge, control or formal authorization of the institutions - practiced by magistrates, civil servants or lawyers who use platforms such as ChatGPT or Claude to draw up decisions or documents without institutional traceability.<\/p>\n A survey by CNJ\/UNDP<\/strong> of 2024 revealed alarming data: over 45%<\/strong> of Brazilian courts and councils reported using generative AI tools. In 57,6%<\/strong> of cases, access is through personal accounts, without institutional centralization. Only 5,9%<\/strong> of agencies have well-defined guidelines for the use of these technologies.<\/p>\n The risks are serious: leaking sensitive data, spreading biases, factual errors (hallucinations) in legal documents and compromising the impartiality of decisions. Civil, administrative and criminal liability for this irregular use is still an issue. gray area<\/strong>, This creates legal uncertainty.<\/p>\nBrazil Builds Infrastructure While the World Speculates<\/h2>\n
Meta Advances on User Data: Trust at Stake<\/h2>\n
Brazilian Managers Use AI for Investment Decisions - With Supervision<\/h2>\n
Risk-Based Regulation: The Road to Reliable AI<\/h2>\n
The Risk of Shadow AI in Law<\/h2>\n