Why a former voice AI company went all in on deepfake detection

Resemble AI, founded in 2019 to develop scalable generative voice technology for gaming, is shifting its primary business from commercial voice AI to multimodal deepfake detection after observing growing misuse of synthetic media. The company argues that its experience building voice models, watermarking tools, and speaker-verification systems provides an advantage in detecting AI-generated audio, video, and images, citing a 2026 report that documented 821 deepfake attacks affecting at least 15,736 people in six months. It will continue supporting existing voice-generation customers and keep models such as Chatterbox and DramaBox open and free for research, but will not sell new commercial voice products. Its new focus is an enterprise “trust stack” of five APIs for detecting manipulated media, embedding and reading watermarks, matching voices and faces to enrolled identities, identifying recurring fraud patterns without storing content, and providing human-readable explanations of detection results. Resemble AI states that its DETECT-World system evaluates media for physical consistency rather than relying solely on known generator signatures, and claims 99.5% audio deepfake detection accuracy on public benchmarks while expanding its research and detection capabilities across all major media formats.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.