PVML
PVML provides a differential-privacy-based secure data access layer that lets AI agents and LLMs query enterprise databases via MCP or API without duplicating or exposing raw sensitive data; it was named a 2025 Gartner Cool Vendor in Data Security.
Visit Website ↗ + Add to CompareOverview
PVML was founded in Tel Aviv in 2022 by Rina Galperin, a Microsoft-alumna NLP/AI specialist, and Shachar Schnapp, who holds a PhD in differential privacy, applying their combined research background to the problem of letting AI systems query sensitive enterprise data safely.
The platform spins up secure, AI-ready ‘virtual databases’ on top of existing infrastructure, applying a differential-privacy engine and permissions enforcement at query execution time rather than requiring data to be copied, moved or pre-anonymized. It auto-generates AI-compatible access protocols (MCP, A2A, API) compatible with tools like ChatGPT and Claude, and maintains centralized audit trails for governance.
PVML raised an $8 million seed round led by NFX in 2024, was named a Gartner Cool Vendor in Data Security for 2025, and lists partners and customers including Intel, NVIDIA, Google Cloud, Accenture and Deloitte, plus a strategic partnership with VisionWave for secure real-time AI in mission-critical operations.
Innovation Matrix Assessment
Moved from differential-privacy research background to a functioning multi-protocol (MCP/A2A/API) product with Gartner recognition within about three years.
Query-time privacy enforcement without data duplication addresses a concrete operational bottleneck security teams face when connecting AI agents to live databases.
An $8M seed round, Gartner Cool Vendor recognition, and named partnerships with Intel, NVIDIA and Google Cloud indicate credible early momentum.
Applying formal differential privacy at query time to gate AI access to live data is a more fundamentally different technical approach than typical rule-based data masking.
Enterprise-name partnerships lend credibility, but independent, published testing of the privacy guarantees under adversarial conditions was not found.
Secure, privacy-preserving data access for AI agents is likely to become more important as agentic AI is connected to more sensitive enterprise data sources.
Why CISOs Should Care
PVML addresses a specific, high-risk gap — how to let AI agents query live enterprise databases for useful answers without duplicating sensitive data or granting overly broad access — a problem that grows more urgent as agentic AI is connected to core business systems.
What Makes It Different
PVML's use of formal differential-privacy techniques (rather than simple masking or rule-based redaction) to gate AI access to live data at query time is a more mathematically rigorous approach than most competing data-access-control products.
The Matrix Verdict
57/100 — INCREMENTAL INNOVATOR
A technically differentiated, research-grounded seed-stage company with real Gartner analyst recognition and credible enterprise partnerships, though as with most companies at this stage independent efficacy validation beyond vendor claims is limited.
Editorial Note: Claims vs. Verified Findings
The Gartner Cool Vendor recognition is independently verifiable; specific claims about zero-duplication data access and query-time privacy guarantees rely on PVML's own technical descriptions, which were not independently audited.
Sources
Alternatives to PVML
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…
Tenzai
An agentic AI penetration testing startup building autonomous 'AI hackers' to find and validate exploitable vulnerabilities at a…
Reco
Reco secures the "agentic ecosystem" — mapping what AI agents can access across SaaS and enterprise apps, detecting…
Charm Security
Agentic AI workforce that investigates and intervenes on scams and fraud in real time, reading manipulation and intent…