*Originally published in Jota.

*This is an AI-powered machine translation of the original text in Portuguese

The Brazilian AI Plan is grounded in the conviction that artificial intelligence is a strategic driver of economic development, productivity, and international competitiveness. At the core of the plan is the promotion of startups, the attraction of investment, and, ultimately, the creation of conditions that enable Brazilian companies to compete in this market.

Across Latin America, AI startups have established themselves as key recipients of venture capital investment, particularly in sectors such as legal tech, fintech, education, and business automation. Brazil is the region’s leading innovation hub, attracting funding from major international venture capital firms and fostering companies with the potential to compete globally.

All indications suggest that this is the country's main bet for advancing in this strategic technology, which requires policy coherence among the various public bodies capable of influencing this agenda. Unfortunately, not all of them appear to be aligned.

At the end of 2024, the Administrative Council for Economic Defense (Cade) exceptionally initiated five proceedings involving investments by Big Tech companies in startups: NVIDIA’s investment in Run:AI, Microsoft’s investments in Mistral and Inflection AI, Google’s investment in Character.AI, and Amazon’s investment in Anthropic. This was exceptional because the startups involved did not meet the minimum revenue thresholds that would make notification of the transactions to Cade mandatory.

Nevertheless, its General Superintendence decided to refer the cases to the Tribunal for review, on the grounds that so-called “digital ecosystems” pose challenges for competition authorities, invoking the principle of in dubio pro societate, but without identifying any concrete elements that would raise competitive concerns.

This approach was explicitly inspired by initiatives undertaken by the UK competition authority and the U.S. Department of Justice, which, in 2024, launched inquiries based on concerns that Big Tech companies could come to dominate artificial intelligence markets through investments in AI startups.

Those authorities referred to acquisitions of startups such as WhatsApp, Instagram, and YouTube by companies such as Facebook and Google, arguing that they had failed to foresee the potentially negative consequences of those transactions for competition. Indeed, those deals have even been described as “killer acquisitions,” on the theory that their purpose was to prevent the emergence of potential competitors.

However, while the UK authority closed its proceedings within six months, Cade took approximately a year and a half merely to decide whether the cases should have been subject to notification. During that period, uncertainty persisted regarding the treatment of partnerships and investments involving AI startups.

Unfortunately, this lack of legal certainty is likely to discourage not only the investments and partnerships of the companies directly involved, but also future transactions involving other startups in the sector, which depend on capital, infrastructure, and strategic cooperation in order to develop and scale.

Far from dispelling these concerns, the Tribunal’s ruling reinforced them: it ordered the notification of the Microsoft/Inflection AI transaction, initiated new proceedings to investigate Google’s investments in the startups Windsurf and Hume AI, and postponed a decision in the Amazon/Anthropic case. The Tribunal made clear that it intends to use its call-in powers to review investments in AI startups even when the objective notification thresholds established by law are not met.

The power to call in transactions for review should be exercised sparingly and only in exceptional circumstances, particularly in a strategic and innovation-intensive sector. For precisely that reason, the rule proposed in Bill No. 4,675/2025—which would require the automatic review of any acquisition or transaction involving technology companies designated as having “systemic relevance” (so-called “gatekeepers”)—is not a welcome development.

There is no comparable mechanism in other jurisdictions. A far more reasonable approach, including to avoid overburdening Cade, would be simply to require the disclosure of essential information regarding such transactions, enabling a prompt decision on whether to exercise call-in powers, without suspending the effects of these growth-enhancing investments.

Otherwise, Brazilian startups—or worse, their potential investors—may be discouraged, as they will face the regulatory risk and cost that partnerships and capital investments could be subjected to prolonged antitrust scrutiny, even where no competitive risk is present. If this is not already a sufficient public policy reason for neither Cade nor Bill No. 4,675/2025 to adopt such an approach, there are also compelling competition-law reasons against it.

First, unlike transactions that occurred in the past in digital services markets, most transactions involving AI startups consist merely of capital investments, sometimes coupled with enabling partnerships, without any acquisition of control or authority over software development or the commercial direction of the startups.

Second, providers of foundation models and AI assistants are competing within the same market, and the partnerships associated with their investments enhance differentiation in ways that stimulate rivalry (see, in this regard, the Legal Wings Institute research report “Competition in AI Markets”). Examples include the integration of AI assistants with search engines, computing capacity, conversational tools, or office software suites.

Anthropic’s recent fundraising rounds, which included investments from major funds in addition to Amazon and global hardware suppliers such as Micron, Samsung, and SK Hynix, have made it the most valuable company in the AI market, surpassing OpenAI. This will undoubtedly strengthen Claude’s ability to compete against ChatGPT.

Third, aside from the fact that the term “killer acquisition” is an unfortunate description of what occurred in so-called digital markets[1]—as no companies or services were shut down—investments by large technology firms in startups generate positive competitive incentives. As Petit and Teece observe, startups often do not seek to replace incumbents but rather to create complementary functionalities.

Partnerships and even acquisitions by Big Tech firms, in turn, lead to increased investment in R&D, access to technological infrastructure, and integration into broader ecosystems, allowing firms to exploit complementarities and develop competitive advantages. For startups, this tends to result in greater scale, faster commercialization, and improved distribution.

Entrepreneurs and venture capital funds, for their part, invest in startups precisely with the expectation of a future exit. There is therefore a positive symbiotic relationship among these actors that tends to foster both innovation and competition.[2]

Accordingly, approaches that increase the regulatory risk associated with transactions involving AI startups may inadvertently harm competition. This is not to deny the importance of antitrust review, but excessive delays can be fatal. If the government seeks to encourage the creation and growth of AI startups, it is inconsistent to signal that strategic alliances with investors or technology partners may be subjected to lengthy and uncertain investigations, thereby increasing the perceived risk of investing in the country.

It therefore seems premature for Cade to create obstacles to partnerships between major technology companies and startups, inadvertently entrenching theories that could ultimately discourage investment in Brazilian companies and, in doing so, undermine policies aimed at fostering innovation and building national technological capabilities.

[1] Marc Ivaldi, Nicolas Petit, and Selçukhan Unekbas, “Killer Acquisitions: Evidence from European Merger Cases,” Antitrust Law Journal, Vol. 86, No. 3 (2024), pp. 647–695.

[2] Nicolas Petit and David J. Teece, “Innovating Big Tech Firms and Competition Policy: Favoring Dynamic over Static Competition,” Industrial and Corporate Change, Vol. 30, No. 5 (October 2021), pp. 1168–1198.