Voxmind
Early-stage London startup building on-device software to flag AI voice deepfakes on live phone calls in real time.
Visit Website ↗ + Add to Compare Claim This CompanyOverview
Voxmind is a London-based startup that raised a £546,491 pre-seed round in 2026, led by Ascension Ventures, to build on-device voice deepfake detection. Co-founded by CEO Jai Keerthi, the company’s pitch is timed to a gap left as larger cloud providers have scaled back standalone voice-biometrics offerings, leaving smaller vendors a narrower but more motivated buyer base in financial services, telecoms, and contact centers.
The product analyzes phoneme-level frequency patterns on-device, without requiring cloud connectivity or GPU acceleration, aiming to flag synthetic or cloned voices within seconds of a live call rather than after the fact. The company says its detection runs in under 500MB of memory, making it deployable on existing call-center infrastructure.
Voxmind is pre-revenue and very early stage; its detection-accuracy figures are currently vendor-reported rather than independently benchmarked, and the funding raised to date is small even by seed-stage standards in this niche.
Innovation Matrix Assessment
The company has a working on-device detection engine and a pre-seed round closed within roughly a year of founding, reasonable early velocity for a pre-revenue team.
Real-time, on-device deepfake flagging on live calls would be operationally useful for fraud-prone call centers, but the product has not yet been shown to work at scale in production environments.
A sub-£1 million pre-seed round with no named enterprise customers is genuinely early-stage momentum, well below what comparable deepfake-detection peers (Pindrop, Resemble AI) have raised.
On-device, GPU-free detection is a meaningful efficiency improvement over cloud-dependent competitors, but voice-deepfake detection itself is an increasingly crowded category, not a new one.
The 99.8% accuracy figure is the company's own claim; there is no independent benchmark, named customer deployment, or third-party test result available yet.
Voice-cloning fraud is a growing, well-documented threat vector, so the underlying problem Voxmind addresses will likely matter more over the next few years even if this specific company does not survive.
Why CISOs Should Care
Offers a lightweight, on-device option for flagging AI-cloned voices on live calls without adding cloud dependency or GPU infrastructure to contact-center systems.
What Makes It Different
Claims on-device, sub-500MB phoneme-frequency analysis rather than the cloud-based, GPU-dependent architectures used by larger voice-biometrics incumbents.
The Matrix Verdict
47/100 — EMERGING / UNRANKED
Emerging/Unranked: a credible early-stage idea addressing a real and growing threat, but with only vendor-reported accuracy claims, no independent validation, and pre-seed-stage funding that leaves genuine market traction unproven.
Editorial Note: Claims vs. Verified Findings
The £546,491 funding round and investor identity are independently reported; the 99.8% detection accuracy and sub-3-second, sub-500MB performance figures are the company's own, unverified by independent testing.
Sources
Alternatives to Voxmind
Adaptive Security
AI-driven platform that simulates deepfake, voice, and multichannel social-engineering attacks to train and test organizations against next-generation phishing.
Quilr
Early-stage agentic AI security startup building a 'Service-as-Software' platform to guard against human-related breaches and secure AI agent…
Zenity
Governance and security platform for AI agents and low-code/no-code development, securing agent identity, permissions and behavior across the…
Gray Swan AI
Carnegie Mellon-linked startup building automated red-teaming and runtime guardrails that frontier AI labs use to stress-test their models…
Tenzai
An agentic AI penetration testing startup building autonomous 'AI hackers' to find and validate exploitable vulnerabilities at a…
Charm Security
Agentic AI workforce that investigates and intervenes on scams and fraud in real time, reading manipulation and intent…