Beware of AI without clear rules: who’s overseeing the algorithm?

Artificial Intelligence is no longer an experimental technology. At home, we read about AI, consult it, question how it is used, and constantly look for ways to optimize its capabilities. Something similar is happening in companies: AI is actively involved in decisions related to credit approvals, fraud detection, pricing, the creation of written, graphic, and audiovisual content, the automation of business processes, customer service, and many other specialized functions.

What is not advancing at the same pace, however, is the discussion around how these decisions are regulated, controlled, and audited. This is happening both within individual organizations and at the national and global levels.

Among the many consequences this brings—ones we will not address in this article—I would like to focus on algorithmic governance as a critical factor in reputation, compliance, and competitive advantage.

Algorithmic governance can be defined as a set of processes, policies, and oversight mechanisms that enable organizations to manage the lifecycle of AI models. If we want and expect these models to influence financial, contractual, or business decisions within an organization, it is important to understand how they work and how they should be governed.

Of course, this issue varies by industry. In sectors such as banking, insurance, and healthcare, for example, an AI model can determine who gets approved for a loan, what level of risk is assigned to a customer, and much more. It is strategically important to understand the logic behind each model, as well as its biases. This includes understanding where the data comes from, its quality, its representativeness, and how it can be monitored to ensure that it meets the standards each organization aims to uphold.

Algorithmic governance is therefore closely linked to regulatory compliance, but it also helps mitigate more subtle risks—those that can emerge before any violation actually occurs. This is why data traceability is so important: organizations need to understand the logic behind the decisions they are automating.

The Latin American Artificial Intelligence Index (ILIA 2025), which analyzed the state of AI in 19 countries across the region by assessing institutional frameworks, international engagement, and regulatory frameworks, highlights the fragmented nature of our region.

There are different realities and outstanding challenges in the adoption process, but the main issue is a gap between stated commitments and actual maturity. As companies seek to scale AI solutions, they encounter strategies without adequate funding, infrastructure bottlenecks, and a lack of global standards.

Countries such as Brazil, Chile, and Uruguay have developed strategies with long-term visions. However, according to ILIA 2025, around seven countries in the region have yet to define a roadmap for AI development. Even among those that have established strategies, few have made significant progress in allocating budgets or developing action plans for effective regulation.

Against this backdrop, and given the lag in government regulation, the World Economic Forum (WEF) suggests that governance should start within the organization. Business leaders cannot afford to wait for specific laws to be enacted; data governance must be treated as a priority.

Faced with the risk of AI autophagy, in which models degrade as they are trained on AI-generated data, companies need to prioritize data provenance systems to ensure that they are building organic, reliable information structures that are integrated into day-to-day operations. The same WEF report mentioned above warns that as the boundaries between real and synthetic data continue to blur, trust will increasingly depend on strong governance.

Viewing trust as a strategic asset is the first step. Companies that can demonstrate how their models work, how they manage controls, and how they mitigate risks will be best positioned to scale with the power of AI. According to the WEF report, trust is becoming the true limiting factor in the expansion of AI, and 81% of companies are still in the early stages of implementing responsible governance practices.

As the region moves toward global standards, Latin American companies have an opportunity to position themselves by creating new avenues for building trust and legitimacy. Governance is both a social issue—one of trust—and a technical one, supported by data centers and infrastructure. Data traceability and provenance systems, along with participatory governance, can help strengthen engagement and legitimacy.

Once trust has been established, organizations can safely leverage synthetic data—data generated by AI models that can be combined with real-world data. Its use can serve as an engine for innovation, enabling companies to scale over the long term.

For example, synthetic data can be used to replicate statistical patterns found in real-world data without exposing personal information. It can also help address underrepresented demographic groups, health conditions, or other factors to improve equity in the financial sector. In this way, organizations can develop models and run simulations to accelerate innovation while reducing the risks associated with using real-world data.

For this advantage to be sustainable, investing in data provenance and traceability systems must be a priority for companies. These capabilities can help mitigate the risk of existing biases being amplified or models degrading over time. If organizations want to scale at speed, they must ensure that data flows as effectively as possible.

A company that implements active transparency mechanisms immediately stands out. Many Latin American companies that sell services to, or are part of, global value chains need to be able to demonstrate how they govern and control the AI systems they use.

A global pharmaceutical company, a leading international bank, or a multinational corporation is unlikely to delegate critical processes to providers that lack mechanisms for explainability and oversight, as well as the ability to detect errors, correct them, and activate mechanisms to address them. Consider Natura and its initiative to democratize information across business areas and suppliers in order to achieve greater scalability without losing sight of governance. The company understands that data has little value if it cannot be traced.

While other competitors may be constrained by data access restrictions or regulatory concerns, these organizations will have a secure platform on which to experiment, train models, and develop new services. Competitive advantage built on trust is the best path toward sustainable growth and genuine innovation—the kind of companies we need for the future.