Skip to content

DataKrypto

Fully homomorphic encryption technology that keeps data and AI models encrypted throughout training, inference, and deployment.

Visit Website ↗ + Add to Compare
65/100Incremental Innovator

Overview

DataKrypto develops fully homomorphic encryption (FHE) that allows computation on data while it remains encrypted, without ever needing to decrypt it. Its flagship FHEnom for AI combines this patented FHE approach with Trusted Execution Environments (TEEs) to create an end-to-end security architecture that protects AI model integrity and data confidentiality across the entire AI lifecycle — ingestion, training, and inference — including while data is actively being computed on in system memory.

Headquartered in Burlingame, California, DataKrypto has achieved FIPS 140-2 validation for its Fully Homomorphic Encryption Module and describes FHEnom as one of the only current homomorphic encryption implementations capable of operating at near-real-time speeds, historically FHE’s biggest practical limitation. The company targets what it calls the “cleartext gap” — the moment data or model weights are typically decrypted for processing — as the last major unaddressed vulnerability in confidential AI computing, with FHEnom for AI now available on Google Cloud Marketplace.

Innovation Matrix Assessment

Innovation Velocity 8/10

Achieved FIPS 140-2 validation and launched on Google Cloud Marketplace in quick succession, unusually fast progress for notoriously slow-to-mature FHE technology.

Operational Value 6/10

Closes the 'cleartext gap' where data is typically decrypted for processing, a real and previously largely unaddressed exposure window in confidential AI computing.

Market Momentum 4/10

Google Cloud Marketplace availability is a positive distribution signal, but funding, customer count, and broader commercial scale remain largely undisclosed.

Category Disruption 8/10

Practical, near-real-time fully homomorphic encryption is a genuinely rare technical achievement that could fundamentally change how confidential AI computing is architected if it holds up at scale.

Real-World Efficacy 5/10

FIPS 140-2 validation is real independent certification for the cryptographic module, but broader performance-at-scale claims remain vendor-published and unverified independently.

Enduring Relevance 8/10

As enterprises push sensitive data into third-party AI models and cloud inference, encrypted-in-use computation addresses a rapidly growing confidentiality concern.

Why CISOs Should Care

Lets organizations use third-party AI/cloud compute on sensitive data without ever exposing it in cleartext, even during active processing.

What Makes It Different

Combines FHE with Trusted Execution Environments to eliminate the cleartext exposure window that most confidential-computing approaches still have.

The Matrix Verdict

65/100 — INCREMENTAL INNOVATOR

A technically ambitious, independently certified deep-tech player tackling one of cryptography's hardest practical problems, still early in commercial scale.

Editorial Note: Claims vs. Verified Findings

FIPS 140-2 validation is independently verifiable via NIST; near-real-time performance claims are vendor-published.

Sources