LC SEC, consulting firm specializing in cybersecurity and compliance, warns of the increasing use of synthetic content in digital scams. Images, videos, and voices produced or manipulated by artificial intelligence have become integrated into strategies for fraud, deceptive advertising, and disinformation. In Brazil, the dissemination of false content created with AI more than tripled between 2024 and 2025, with a growth of 308%, according to the first Panorama of Disinformation in Brazil, by the Lupa Observatory, which maps trends, targets, and main tactics of disinformation. This scenario is also analyzed by the National Data Protection Authority (ANPD) in Technological Radar No. 6 – Deepfakes, a 163-page study detailing how synthetic content is being used in different types of scams and rights violations.
One of the points of concern highlighted by the authority is the use of deepfakes in sponsored ads and boosted campaigns on social media. According to the document, criminals can use the image or voice of well-known people to create fake offers, promote content to segmented audiences, and direct victims to cloned websites or fraudulent platforms where payments or personal information are requested.
The ANPD's publication comes at a time of advancing discussions about the responsibilities of digital platforms and the need for mechanisms capable of reducing the circulation of illicit content. The report also addresses applications of deepfakes related to financial scams, misleading advertising, electoral disinformation, gender-based political violence, racism, and non-consensual use of images, showing that the risks go beyond the field of cybersecurity and reach issues of privacy, reputation, and trust.
One of the Brazilian cases analyzed by the authority involves deepfakes of celebrities used to promote non-existent products. According to the report, the investigation of the responsible gang identified transactions exceeding R$ 20 million. The Brazilian National Data Protection Authority (ANPD) also draws attention to the ability of these scams to operate on a large scale, with relatively small individual losses that, when added together, can generate high profitability for criminals.
The FBI, for example, received 22,364 complaints related to the use of artificial intelligence in fraud in 2025, with adjusted losses exceeding US$893.3 million. Companies reported more than US$30 million in losses related to corporate email compromise scams that used some AI component. The World Economic Forum's Global Cybersecurity Outlook 2026 indicates that 73% of respondents were directly affected or knew someone affected by digital fraud in 2025, while 87% perceived an increase in vulnerabilities associated with AI.
For Luiz Claudio, CEO and founder of LC SEC, the advancement of deepfakes requires companies to treat identity and authentication as security elements that cannot depend solely on a person's appearance or voice. “The problem is that traditional identity validation processes were built in a scenario where voice, image, and presence in a video conference were relatively reliable signals. With the evolution of artificial intelligence, these elements can be reproduced. Companies need to create independent confirmation mechanisms for critical decisions, especially when they involve money, access to systems, or sensitive information,” he states.
Digital advertising and the financial sector are among the areas most exposed to this type of fraud. In addition to verifying the origin and authorization for the use of images and voices, companies need to adopt additional identity confirmation mechanisms in sensitive operations. The risk became evident in a case involving Arup, when criminals used synthetic representations of executives in a video conference and convinced an employee to transfer approximately US$25 million.
According to the expert, the adoption of additional layers of verification is fundamental to prevent synthetic content from being used as the sole form of authentication in sensitive operations. “Organizations can adopt confirmation mechanisms through independent channels for higher-risk operations. Payment requests, changes to bank details, document sharing, or granting access may require a second validation through a previously known channel, in addition to double approval, segregation of duties, strong authentication, and audit logs. The combination of these controls reduces the possibility that a single synthetic representation will be sufficient to authorize a critical action,” highlights Luiz Claudio.
Decrees No. 12,975 and No. 12,976, published in May 2026, expanded the ANPD's competencies regarding governance, transparency, fraud prevention, and illicit content on platforms. In the electoral environment, TSE Resolution No. 23,755/2026 also established rules for the use of synthetic content in advertising, including identification requirements and restrictions on the use of images and voices of candidates and public figures.
According to Luiz Claudio, combating deepfakes requires a protection strategy that combines technology, internal processes, and employee awareness. "The technology for generating synthetic content evolves rapidly, and it's unrealistic to rely on a single tool to identify all deepfakes. Protection needs to be distributed across people, processes, and technology. When a company establishes clear rules for sensitive decisions, monitors the misuse of its brand, and prepares employees to question requests outside of usual channels, it creates another layer of protection against this type of fraud," he concludes.
