ScamAI is an AI-powered platform for detecting deepfakes and synthetic media in real-time. We provide AI detection, audio detection, and GenAI content verification with industry-leading accuracy.

How accurate is ScamAI's deepfake detection?

ScamAI's Eva-v1 models are continuously trained on the latest synthetic media techniques with processing times under 4 seconds. Accuracy varies by media type and attack technique — contact us for benchmark details relevant to your use case.

Is ScamAI compliant with security standards?

Yes, ScamAI is SOC 2 Type II compliant and GDPR compliant, ensuring enterprise-grade security and data protection.

How much does ScamAI cost?

ScamAI offers 200 free images per month for free. Pay-as-you-go pricing starts at $0.05 per image. Optional add-ons (face attributes, liveness detection, ID document verification) are $0.008 per image each. Enterprise pricing is available for high-volume users.

How does ScamAI compare to other deepfake detection tools?

ScamAI offers both vision and audio detection in a single platform with sub-4-second processing, and provides a developer-friendly API with integration in under 10 minutes. Unlike many competitors, ScamAI offers a free tier of 200 images per month, transparent per-image pricing, and is SOC 2 Type II and GDPR compliant.

What industries use ScamAI?

ScamAI is used across financial services for KYC verification, content platforms for moderation, call centers for voice authentication, media organizations for content verification, HR departments for remote interview verification, insurance for claims validation, and government agencies for public sector security.

Document ForgeryGPT-Image-2Benchmark

When the Forger Is the Judge: GPT-Image-2 Cannot Recognize Its Own Faked Documents

Jiaqi Wu, Yuchen Zhou, Dennis Tsang Ng, Xingyu Shen, Kidus Zewde, Ankit Raj, Tommy Duong, Simiao Ren

Read on arXiv

How to cite

@article{wu2026forger,
  title={When the Forger Is the Judge: GPT-Image-2 Cannot Recognize Its Own Faked Documents},
  author={Wu, Jiaqi and Zhou, Yuchen and Ng, Dennis Tsang and Shen, Xingyu and Zewde, Kidus and Raj, Ankit and Duong, Tommy and Ren, Simiao},
  journal={arXiv preprint arXiv:2604.25213},
  year={2026}
}

Associated datasets

AIForge-Doc v2.0 (GPT-Image-2 document forgeries)

A v2 expansion of AIForge-Doc covering GPT-Image-2 generated tampering — used in the 'When the Forger Is the Judge' benchmark.

When the Forger Is the Judge: GPT-Image-2 Cannot Recognize Its Own Faked Documents

How to cite

Associated datasets

More from Document Forgery