Resemble has always treated misuse as a core problem. When we built generative voice models, safeguards shipped with it. Now that we’re focused entirely on detection, the tools we build to determine what’s trustworthy have to earn that trust as well.
Detection was part of how we built long before it became our focus. Our voice work required explicit consent and speaker verification before any clone could be created. We developed imperceptible watermarking so generated audio could be traced to its source. We open-sourced Resemblyzer so researchers could analyze and compare voices. And we built our first deepfake detector to catch attempts to misuse our own platform. Those safeguards became the foundation for everything we build now.
Consent for identity — Resemble Identity only enrolls a voice or likeness with the explicit consent of the person it belongs to.
Your data stays yours — Resemble Meetings does not record or transcribe your meeting in the process of detection to keep conversations confidential.
Detection informs, people decide — Results provide context for a decision. Your team, enterprise policies and procedures drive the next steps.
Honest about performance — We advocate for and publish third-party benchmarks and we’re clear that no detector is perfect. Even when we achieve 100% accuracy, we claim below that to allow for the possibility of an error down the road.
Provenance that’s not performative — Our PerTh Multimodal model isn’t just for detecting the existence of a watermark but is also for creating one and for complying with C2PA requirements to make origin verifiable, not assumed. PerTh will not apply a watermark to media that already carries a mark so it can’t be overwritten.
Enterprise-grade security — Resemble is fully compliant with GDPR, SOC 2 Type II, HIPAA, and is internationally recognized information security management systems.
Resemble products may not be used to:
We contribute open-source tools, publish our benchmarks and research, track real-world harm in our Deepfake Incident Database, and offer detection access to journalists through our Press Access Tier. We engage with industry, standards bodies, and policymakers, from C2PA to the EU AI Act and California’s SB 942, to help shape how AI accountability works in practice. And we’re the only detection company to sign the EU’s Code of Practice on Transparency of AI-Generated Content, as a technology provider of marking and detection solutions.