By: Alex Mercer – SeaPRwire – Most detection tools still examine one media type at a time. Scammers stopped doing that years ago. A cloned voice on the phone sets the urgency. A fabricated image or document follows. A manipulated video seals the ask. Scam.ai and Modulate just announced a partnership that treats the whole sequence as one problem. On August 4 they said Modulate’s synthetic voice models will sit inside the Scam.ai platform. Customers will analyze image, video, and audio through a single workflow.
The official facts are precise. Scam.ai already covers images, videos, and digital documents. Modulate brings specialized synthetic voice detection. The combined system returns confidence scores and detection signals for all three. Dr. Ben (Simiao) Ren, Scam.ai co-founder and CEO, said scammers left the single-channel approach long ago while many detection systems stayed organized by media format. The integration lets customers add voice detection to the same platform and workflows they already use for visual content. Carter Huffman, Modulate CTO and co-founder, noted that voice now forms part of coordinated multi-media scams. A convincing clone builds trust. Fabricated visuals reinforce it. Modulate’s model reports 98.9 percent accuracy and a 1.1 percent equal error rate. As of August 4 it held first place on the Hugging Face Speech Deepfake Detection Leaderboard. It handles real-time streaming and prerecorded audio. Scam.ai’s Eva-v1 models report 98.2 percent visual detection accuracy against the company’s internal benchmark. Both expose results through APIs built for enterprise integration. The joint capability is expected in early September.
The quieter commercial point is consolidation. Gallup and the Stop Scams Alliance estimated 15.1 million U.S. adults were personally scammed in 2025, with losses of at least 68 billion dollars. Phone calls, text messages, and email each appeared in 45 percent of scams. Half of the incidents crossed two or more communication methods. Phone calls ranked as the primary channel more often than any other. Defenses that examine only one piece of the interaction miss the pattern. Organizations that already run Scam.ai for visual checks can now fold voice into the same interface. They avoid standing up another standalone tool. Potential uses listed in the announcement include identity verification, payment authorization, executive impersonation, contact-center security, content moderation, insurance claims, digital evidence, and enterprise investigations. Confidence scores let teams prioritize high-risk items for human review.
The pattern for security buyers is already clear. Single-format detectors will keep losing ground as attacks chain channels. Platforms that deliver image, video, and voice scores inside one workflow will own the next round of procurement conversations. The practical test for any fraud or security team is simple. Run a real multi-channel sample through the combined system once it ships in September. Measure whether the joint signals catch sequences that separate tools miss. If they do, the partnership has closed a gap that cost real money in 2025. Author bio: Alex Mercer, a technology director and analyst who has spent years inside large-scale engineering organizations evaluating how detection systems perform against coordinated synthetic-media attacks.
source https://newsroom.seaprwire.com/press-releases/technologies/deepfake-scams-already-cross-channels-scam-ai-and-modulate-just-stopped-treating-them-as-separate-problems/