AI Mistrust, Rising Fraud Expose a Dual Digital Trust Crisis

Organizations face a dual trust challenge in the digital era. On one hand, 69% of decisions based on AI require human validation due to governance gaps, according to Dynatrace. On the other, the Mexican financial system registered 2.48 million fraud claims in 1H25, according to CONDUSEF, exposing critical operational security failures.

The problem splits into two fronts that, while distinct, reveal a gap between technology adoption and control. Internally, scaling AI depends on new supervision mechanisms. "Observability becomes the shared intelligence that allows [for] smarter and safer decisions at scale," says Alois Reitbauer, Technology Strategist, Dynatrace. Externally, "users are more connected, but not necessarily more protected."

The internal challenge is defined by the need to scale AI reliably. The Dynatrace 2025 State of Observability Report indicates that while 100% of surveyed business leaders use AI, mistrust in the models compels the majority (69%) to maintain manual review processes before executing critical decisions.

This operational friction has motivated a direct financial response. Seventy percent of organizations increased their observability budgets during the last year. Furthermore, 75% plan to increase this investment again in the next fiscal cycle. The objective is to ensure that AI-based systems are transparent, reliable, and efficient. This allows their application in high-value areas like data management (57%), AI governance (50%), and security operations (46%).

Concurrent with this internal governance struggle, organizations face an external trust crisis from tangible security failures. A report from Mexico's National Commission for the Protection and Defense of Users of Financial Services (CONDUSEF) regarding 1H25 quantifies the impact of banking fraud in Mexico.

User claims for potential fraud increased 5.2% year-over-year in Mexico, reaching 2.48 million cases. The total amount claimed by affected parties rose to MX$10.71 billion (US$580 million). However, the banking sector's response and compensation capacity is limited. Institutions only reimbursed MX$2.56 billion (US$138 million), which represents a reimbursement rate of just 25% of the claimed amount.

A deeper look into the CONDUSEF data reveals the predominance of the digital vector in these attacks. Over 60% of complaints originate from digital fraud. The most common methodologies include phishing, fraudulent calls, and deceptive applications that impersonate institutions.

Identity theft is consolidating as one of the fastest-growing modalities. Claims for impersonation totaled MX$634 million during the period, a 24.6% annual increase. In this category, the institutional response is even more deficient, with reimbursements reaching only MX$65 million. This is equivalent to just 10% of the claimed amount. The phenomenon also affects development banks; Banco del Bienestar accumulated claims of MX$43.08 million, a 27.5% year-over-year increase.

By 2030, 50% of companies are expected to adopt AI-based encryption, risk assessment, and threat detection, which will only be viable if current vulnerabilities are resolved, says Dynatrace. Organizations must therefore invest simultaneously in the governance of their internal systems (AI observability) and in the protection of their external ecosystems (fraud defense), as failure on either front inhibits the viability of the digital economy.