Kizwork flags cloned and synthetic voices live, inside the meeting — 99.1% accuracy, alert in 2–3 seconds — and certifies who is speaking. Before the wire goes out, not after.
Voice cloning is now cheap, fast and convincing. The attack no longer targets your network: it targets the person who approves the payment.
An earnings call, a podcast or a conference talk gives a fraudster enough public audio to clone an executive's voice.
An employee in Hong Kong transferred about US$25 million after a video call in which the CFO and colleagues were deepfakes.
In May 2025 the FBI warned that AI-generated voice messages were impersonating senior US officials to gain trust and access.
Every voice in the meeting is analysed live and shown as human or synthetic: 99.1% accuracy, alert in 2–3 seconds (balanced test set, 50% real / 50% synthetic voices), 3% equal error rate on audio degraded down to 10 dB. Independent validation commissioned.
Certified voice identity: a 25-sentence enrolment with ID verification. In our tests the owner is recognised 100% of the time and an impostor is caught 95.5% of the time (target ≥ 98%).
The alert appears in the meeting itself, and a voice challenge (2FA) lets the real person confirm who they are — so the approval can wait for proof instead of trust.
Urgent transfer requests, new supplier bank details, “confidential” acquisitions: the classic scenario of CEO fraud, now with the right voice.
Strategy, M&A and crisis meetings where one fake participant can extract decisions or information.
Video interviews where the candidate's face or voice is not their own — the entry point for insider access.
No need to move your meetings to a foreign cloud. The same detection and identity technology runs in three ways.
AWS, Microsoft Azure, Google Cloud or another provider, as a private deployment inside your own environment.
On your own servers (Docker/Kubernetes), air-gap capable — GPU, or CPU and ARM without a GPU card. Recordings and transcripts never leave your perimeter.
Hosted in the European Union on Scaleway, for teams that want nothing to install and data outside US jurisdiction.
Detection and identity come from CandyVoice, our exclusive partner — MUVOC® patents, ICCA 2026 award winner.
Speech-to-speech translation in each speaker's own voice, with no third-party AI voice service in the loop.
Designed under the GDPR and the EU AI Act: every voice we synthesise is marked as synthetic, with the speaker's explicit consent.
Not yet, and we prefer to say so. Our certification path is ISO 27001, then ISO 42001. Our security architecture, test results and data-processing terms are available to your security team on request.
No. You can run Kizwork in your own AWS, Azure or Google Cloud environment, or on-premises. The Kizwork cloud is in the EU; in your own cloud, the jurisdiction is that of your provider.
Deepfake detection is included in Pro Security, from €14.99 per user per month (excl. VAT), from a single seat. The free demo lets you see detection work first. Private deployments in your cloud or on-premises are quoted on Enterprise Fortress. See pricing.
A 30-minute private demo on your scenario: every voice in the meeting marked as human or synthetic, live. Meet us at Web Summit Lisbon (ALPHA programme), or book a call.
Book a private demo Try the free demo