By Juliana Bauer Lomonaco Quinto, Marketing Manager at Access for Latam
Brazil recorded almost 7 million attempts at digital fraud in the first half of 2025 alone, which represents an attempt every 2.3 seconds, according to data from the Serasa Experian Fraud Attempts Indicator.The number is record and reflects a high of 29.5% over the same period of the previous year. The escalation is no coincidence: the movement accompanies the massification of digital financial services, the consolidation of Pix as the main means of payment and, above all, the arrival of generative artificial intelligence tools to the arsenal of fraudsters.
Companies of all sizes in the financial sector are being forced to abandon traditional document verification models and adopt more robust architectures, with the use of automation and AI. The banking and card sector concentrates more than half of all occurrences involving digital crimes, accounting for 53.7% of attempts in the first half of 2025. The potential loss, that is, the amount that would be lost if fraud were not stopped, reached R$ 39.8 billion in the period, against R$ 29.9 billion in the corresponding interval of 2024.
According to the ClearSale Fraud Map 2024, there were 2.8 million attempts in Brazilian e-commerce, with R$ 3 billion in potentially lost values and fraudulent average ticket of R$ 1,072. Cell phones, home appliances and computers are the most targeted segments.
Generative artificial intelligence as a fake document factory
The most significant change in this cycle of threats is the role of generative artificial intelligence as a fraud industrialization tool. Sumsub points out that Brazil concentrates 39% of all deepfakes detected in Latin America, with incidence five times higher than that of the United States. In the first quarter of 2025, deepfake attacks grew 700% compared to the same period of the previous year, and attempts with synthetic documents advanced 195%.
According to the Entrust Identity Fraud Report 2025, digital document fraud has for the first time surpassed physical forgery, accounting for 57% of global cases.One in 50 falsified documents identified in 2025 was generated by generative AI tools.
The 34th FEBRABAN Banking Technology Survey, conducted by Deloitte, points out that Brazilian banks planned to invest, last year, a higher number 13% compared to 2024. About 10% of the banking IT budget is directed to cybersecurity and fraud prevention, and 100% of the banks interviewed classify the topic as a strategic priority. Investments in IA, analytics and big data are expected to grow 61% in the year, with 80% of fraud detection and laundering applications concentrated.
Document validation, specifically, has become a more complex architecture, which the market calls IDP, acronym for Intelligent Document Processing. The technology combines image capture and pre-processing, automatic classification with machine learning, extraction by intelligent OCR, cross-checking in official databases such as IRS, Senatran and TSE, as well as facial biometrics with proof of active or passive life and forensic analysis of metadata.
Studies by global technology consultancies indicate cost reductions of up to 20% and faster 34% processing with the integrated adoption of these layers.
In biometrics, NIST tests show failure of only 0.2% in the top algorithms, while Brazilian digital identity validation solutions report accuracy of 99.9%, processing tens of millions of monthly validations. Defense logic, however, migrated from isolated biometrics to what experts call a multilayer approach: customer identity is continuously validated throughout the relationship, crossing device data, navigation behavior, geolocation and transactional history with the documentary and biometric information collected at the entrance.
From backoffice to security architecture
What many companies have not yet realized, however, is that validating a document at the gateway is not enough when the internal life cycle of this document is fragile. Once approved, the contract, proof of income or power of attorney must travel through a traceable flow, with immutable audit trails, granular control of permissions, versioning and automated retention policies.
When a tampered document encounters an internal process without traceability, the single point of failure is not the verification technology at the entrance, it is the absence of governance along the entire path.
This is why document management has come to be treated as a security layer, and no longer as a backoffice function. The integration between identity validation platforms, document process automation and electronic document management systems creates an environment in which each step of the flow acts as an additional barrier against fraud.
Document validation is no longer a bureaucratic step of onboarding and has become a piece of security architecture. The combination of stricter regulation, fraudsters armed with generative AI and increasingly exposed consumers imposes a response that is not solved with a single tool. It requires layers of technology, information governance throughout the entire lifecycle of documents and continuous adaptation capacity. Who still treats document verification as punctual checking is, in practice, operating with the door open.


