DuckDuckGoose has published a threat intelligence report warning that AI-generated synthetic identities are bypassing digital onboarding systems at scale across Latin America's financial sector.
The findings indicate that identity fraud has moved beyond the manipulation of existing documents or images toward the real-time generation of entirely synthetic human identities capable of clearing standard onboarding and authentication controls.
The report documents a rapid acceleration in the production of synthetic media tooling. During Q4 2025 alone, more than 55 synthetic media generators were publicly released, approximately one every 1.6 days. Since early 2024, image-to-video generation capabilities have grown by over 1.000%, a shift that directly affects the reliability of biometric checks and liveness detection, two controls widely used in digital financial onboarding.
Moreover, monitoring of open AI ecosystems identified close to 868.000 fine-tuned synthetic identity variants being created monthly. Each variant introduces characteristics that verification systems have not previously encountered, creating windows of exposure during which fraudulent identities may pass controls designed against earlier fraud typologies.
The report frames this as a structural challenge rather than a cyclical one. As generator release velocity increases, detection models face a persistent gap, verification environments must contend with a continuous stream of novel synthetic profiles rather than a fixed set of known attack patterns.
The report also focuses specifically on Brazil and the wider Latin American market, where conditions have amplified the risk. The rapid adoption of PIX, Brazil's real-time payments infrastructure, together with large-scale remote onboarding and neobank growth, has placed identity verification at the centre of financial control frameworks.
Brazilian institutions recorded USD 1.9 billion (BRL 10.1 billion) in banking fraud losses in 2024, and fraud execution timelines are increasingly measured in minutes rather than days. These dynamics heighten the operational consequences of any gap between synthetic identity generation capability and detection infrastructure.
Implications for identity infrastructure
The report's central conclusion is that conventional KYC and liveness verification controls remain functional as designed, but were not built to handle identities that are continuously generated and adapted faster than detection models can be updated. Furthermore, the underlying assumption of such controls no longer holds in an environment where generative AI can produce synthetic presence in real time.
A company official cited in the report stated that trust must be established at the moment of identity creation rather than following fraud detection. For identity platforms operating in digital economies such as Brazil, the report argues that synthetic media detection is transitioning from an advanced capability to a required infrastructure control, positioned alongside document verification and liveness checks as a baseline component of the onboarding stack.
