Cases involving fraud in the São Paulo court system have increased steadily in recent years and have more than doubled since generative artificial intelligence platforms became widely available at the end of 2022. The data comes from a survey conducted by the legal platform Jusbrasil at the request of Folha de S.Paulo. Fraud cases include scams carried out through Pix, Brazil’s instant payment system, a category of crime often referred to by the Federal Police as “digital banditry.” According to the survey, the number of decisions issued by the São Paulo State Court of Justice (TJ-SP) in fraud-related cases rose from 1,073 to 2,270 in 2025, reflecting a trend similar to the increase in police reports involving banking fraud and digital scams.
This Content Is Only For Subscribers
To unlock this content, subscribe to INTERLIRA Reports.
Fraudulent Campaigns
Several factors have contributed to the growing sophistication of fraud schemes. Among the main indicators are the proliferation of large-scale fraudulent campaigns, facilitated by AI-generated content; the disappearance of spelling and grammatical errors that traditionally helped identify scams; the growing participation of foreign criminal groups operating in Brazil; the use of deepfake technology to imitate voices and appearances; and increasingly advanced programming techniques. Investigations reviewed by the report reveal that criminal groups have used artificial intelligence to attempt to bypass facial-recognition systems in banking applications, create fraudulent advertisements, clone voices, and impersonate public figures in so-called “romance scams,” in which victims are manipulated into believing they are engaged in a genuine emotional relationship before being extorted for money.
Limited Data
Jusbrasil notes that an analysis of decisions issued in 2026 concerning cases filed between 2016 and 2026 suggests a growing backlog of unresolved fraud-related proceedings. However, the available information in many lawsuits remains limited. Of the 8,338 fraud-related decisions issued by the TJ-SP in 2026, a total of 3,786 did not specify whether the fraud occurred in person or through digital means. Among the cases where this distinction was identified, 3,621 involved electronic fraud and 931 involved in-person schemes. The use of artificial intelligence is rarely mentioned in judicial proceedings and appeared explicitly in only five cases. This contrasts sharply with broader industry data. According to a survey conducted by CrowdStrike, the use of AI in fraud schemes worldwide increased by 89% and is now associated with nearly 70% of criminal operations.
Bank Accountability
According to Alessandro Barreto, a Civil Police investigator from the state of Piauí, the data reflects two parallel realities. He argues that many victims prioritize seeking reimbursement through legal action against financial institutions rather than reporting crimes to law enforcement authorities. “Once the person receives compensation, they often consider the issue resolved,” he explains. Barreto notes that although both the Civil Police and Federal Police have expanded operations targeting cybercrime in recent years, only a small portion of reported incidents lead to full investigations, largely due to difficulties in obtaining digital evidence and identifying suspects. According to the Brazilian Public Security Forum, law enforcement agencies in Brazil successfully solve only 2% of fraud cases.
New Law
In an effort to curb cybercrime, the federal government enacted new legislation in early May, increasing penalties for fraud and theft committed through digital means. Despite the tougher legal framework, the Ministry of Justice and Public Security acknowledges that it lacks comprehensive data distinguishing whether fraud offenses occur in physical or digital environments, a limitation that mirrors the information gaps identified in court records. There is also no centralized database tracking the use of artificial intelligence in fraud schemes. The federal government currently concentrates reports related to banking fraud through the National Alliance to Combat Bank Fraud, a cooperative initiative involving the Ministry of Justice and the Brazilian Federation of Banks (Febraban).
Court records provide valuable insights into the infrastructure commonly used by fraudsters. WhatsApp is the platform most frequently mentioned in fraud-related proceedings, appearing in approximately 1,300 cases. It is followed by internet banking systems (752 mentions), telephone calls (405), fraudulent websites (270), and e-commerce platforms (146). The cases also reveal the most common criminal tactics. Fake sales schemes rank first, with 416 recorded cases, including situations in which buyers transferred money to fraudsters or sellers were deceived into shipping products without receiving payment. Fraudulent call centers, where criminals impersonate bank employees to obtain sensitive information from victims, appear as the second most frequently reported scam method.
Analysis:
The rapid growth of fraud cases in São Paulo reflects a broader transformation in the criminal landscape, where technological innovation is increasingly benefiting offenders faster than institutions can adapt. The expansion of generative artificial intelligence has reduced many of the traditional indicators that once helped victims identify scams, such as poor grammar, unrealistic messages, or poorly edited images. Criminal groups can now produce convincing content at scale, personalize attacks, and automate parts of the fraud process. As a result, fraud is evolving from a relatively low-cost criminal activity into a sophisticated business model capable of reaching thousands of potential victims simultaneously with a higher probability of success.
The data also exposes a significant gap between the current scale of cybercrime and the capacity of law enforcement and judicial institutions to measure and investigate it. The fact that thousands of court decisions fail to specify whether fraud occurred through digital or physical means illustrates the limitations of existing databases and reporting systems. This lack of detailed information makes it more difficult to identify trends, allocate resources, and formulate effective public policies.
Sources: A Folha de SP.
