Seon expands AI signals to detect synthetic fraud

Seon expanded its signal intelligence to detect and prevent fraud created or assisted by artificial intelligence. The company said the upgrade adds models and analysis to its existing device, behavior and network data.

Those signals are processed to identify patterns and artifacts common to AI-created content and automated account-creation techniques. The results feed into Seon’s risk engine to enable faster scoring and action within merchants’ existing fraud workflows.

Seon provides fraud-prevention software to online merchants, fintech firms, marketplaces and gaming platforms. The enhanced signals are available through Seon’s API and dashboard, allowing clients to block, challenge or accept transactions based on scores that include the new AI-detection inputs.

The upgrade works with Seon’s existing device fingerprinting, IP and velocity checks rather than replace them. The new layer adds classifiers for content artifacts and automation fingerprints and exposes those results in the same risk-scoring framework.

The company said the update targets synthetic-account creation using AI-generated identities and photos, phishing and scam messages written by language models, voice-cloned social-engineering attempts, and automated bots that mimic human browsing patterns. Seon combines behavioral telemetry such as typing and navigation patterns with metadata and content signals to flag likely AI involvement.

Clients can adjust thresholds for the new signals, review flagged cases on the platform, and send confirmed incidents back to Seon to refine detection models. Seon plans to roll the enhancement into its existing product tiers so customers can adopt the AI-focused signals incrementally while monitoring approval rates and chargeback volumes.

Security vendors have added automated detection for synthetic media and machine-crafted content as generative AI tools become more widely available. Seon described the signals as additional evidence points for risk decisions rather than standalone proof.