Banxico Puts AI Under the Microscope: Financial Efficiency Comes With New Red Flags for Mexico
The central bank warns that the rapid adoption of AI could amplify fraud, opacity, and operational risks across banks, fintechs, and Sofipos.
The Bank of Mexico (Banxico) has raised its alert level regarding the use of artificial intelligence (AI) in Mexico’s financial system. While it acknowledges that these tools can improve decision-making, cut costs, and expand access to credit, it also stresses that AI can make episodes of financial stress faster and harder to contain, while complicating oversight due to its lack of transparency.
In its analysis, “Artificial intelligence and financial stability: potential channels for risk transmission,” Banxico argues that AI does not always create entirely new risks, but it does change their scale and the speed at which they spread. In a system like Mexico’s—where large banks with robust infrastructure coexist with a growing ecosystem of fintechs, Sofipos, and other institutions with varying levels of technological maturity—that feature increases the complexity of managing credit, operational, liquidity, and reputational risks.
AI adoption is accelerating in areas such as credit origination, fraud detection, customer service, and product personalization. In particular, it has gained relevance in alternative models used to assess people without a formal credit history—ranging from consumption patterns to job stability. For a country with high levels of informality and gaps in financial inclusion, these technologies promise to broaden financial access; however, Banxico warns that the upside may come with vulnerabilities if models are not verifiable, auditable, or consistent.
The central bank also emphasized a regulatory asymmetry: nonbank institutions, operating under frameworks that are generally less stringent than those applied to banks, may be able to deploy AI models more quickly to compete—but also with uneven controls. In practice, this pushes the entire sector to innovate faster, increasing the risk that solutions are implemented without sufficient data governance, without robust testing, or without clarity on accountability when an algorithm fails.
There is also a financial dimension. Banxico documents notable growth in private credit worldwide directed toward AI-linked companies, and estimates that the infrastructure rollout will require investments close to $3 trillion by 2028, with a significant portion financed through debt. In an environment of still-restrictive interest rates and lower risk appetite in international markets, that leverage can become a channel of contagion if tech projects fail to deliver expected cash flows or refinancing conditions worsen—affecting critical providers and, by extension, financial institutions that use their services.
Cybersecurity, deepfakes, and tech concentration: the Achilles’ heel
Banxico identifies that the development and provision of AI capabilities is concentrated among a small number of global infrastructure and computing providers, increasing the financial system’s dependence on shared platforms and raising the risk of “single points of failure.” For Mexico—where the digitization of payments and financial services continues to advance steadily—this risk is especially sensitive: a technical disruption or shared vulnerability can escalate from an isolated incident into a systemic event, affecting operational continuity, clearing and settlement, and customer service.
The threat is not only technical, but also criminal. The report cites international estimates pointing to a sharp uptick in fraud enabled by generative AI, including deepfakes, which make social-engineering attacks cheaper and easier to scale. Banxico reported that between January and May 2026 it detected eight cyber breaches against financial institutions in Mexico—three banks, two Sofipos, one fintech, and one cooperative—with an uptick in May. That data reinforces the view that technology adoption is happening in parallel with a more professionalized fraud ecosystem, forcing stronger monitoring, authentication, and incident-response capabilities.
On the economic front, the message arrives at a time when Mexico is seeking to strengthen credit to households and small and medium-sized businesses as a growth lever, while facing structural challenges such as low productivity and high informality. AI’s promise to lower origination costs and improve risk assessment could support that agenda; however, if models introduce bias or make systematic errors, they can restrict credit for certain groups or create a false sense of security in portfolios, affecting delinquency and market confidence.
Banxico also suggests that supervision faces a methodological challenge: “black box” models that are not standardized and are difficult to explain complicate audits and accountability. For end users, this translates into a real risk: opaque automated decisions on credit approvals, limits, rates, or preventive account blocks can erode perceptions of fairness and open the door to complaints. For institutions, reputational risk can be immediate if a case of algorithmic discrimination or large-scale fraud goes viral.
Looking ahead, the challenge for Mexico’s financial sector will be balancing innovation with controls: stronger data governance, robustness and bias testing, decision traceability, and technology continuity plans. At the same time, the rise in cyber incidents suggests that investment in security—from strong authentication to transaction monitoring and staff training—will be a competitiveness factor, not just a compliance requirement.
In short, Banxico recognizes AI’s potential to expand access and improve efficiency, but warns that its opacity, its reliance on a concentrated set of providers, and its use by fraudsters can amplify risks. The task will be to innovate without undermining stability and trust in the financial system.
