In today’s digital economy, data is the most valuable commodity—and yet, paradoxically, it remains the least protected. While enterprises spend heavily on firewalls, identity controls, and endpoint protection, the data itself often sits exposed, accessible to anyone with system privileges. This is the exact problem Luigi Caramico, Founder and CTO of DataKrypto, set out to solve.

In a recent interview on An Eye On AI hosted by Swapnil Bhartiya, Caramico broke down how DataKrypto’s unique approach to encryption is setting new standards for data protection in the AI age.

The Problem: A False Sense of Security

Caramico uses a powerful analogy: Imagine a bank with steel walls, cameras, alarms—and unprotected stacks of cash. In cybersecurity, we’ve built thick walls but left the cash (data) vulnerable. Admins and insiders with access can still extract sensitive information. This is not just a theoretical risk; it’s a daily threat across enterprises globally.

The Solution: Continuous Encryption with FHE

DataKrypto implements continuous encryption, ensuring data is encrypted at creation and stays encrypted until a human explicitly needs to consume it. The breakthrough? Using fully homomorphic encryption (FHE)—a method that allows computation on encrypted data without needing to decrypt it.

This is a game-changer for AI and big data workflows. With FHE, systems can perform calculations, searches, even image processing—all while the data remains in its encrypted state.

Performance without Compromise

One of the key differentiators for DataKrypto is its use of symmetric encryption rather than RSA’s traditional asymmetric models. Symmetric keys are smaller, faster, and offer higher security levels in many scenarios. This enables DataKrypto’s platform to operate nearly at plaintext speed—processing encrypted data with minimal latency.

Caramico explains: “We can process data like images, audio, and video at nearly zero delay. For strings, we’re only nanoseconds slower due to slight data inflation.”

This is a significant advantage, especially for real-time applications and high-throughput AI workloads.

AI Security: A New Battleground

As enterprises increasingly rely on AI models to house proprietary data and business logic, protecting those models has become vital. Caramico points out that AI has become the de facto database—and therefore, the #1 target for cybercriminals.

DataKrypto offers protections for both the AI model and the user’s inputs and outputs, ensuring end-to-end privacy. Even the AI provider can’t see the data exchanged with the model.

Using Trusted Execution Environments (TEEs) combined with FHE, DataKrypto isolates the model, encrypts the communications, and decouples the access layers. This makes data theft virtually impossible without physical access to the hardware itself.

A Developer-Friendly Future

Developers integrating DataKrypto’s system into AI workflows will need to plan hardware deployment using TEEs, but Caramico assures that the performance impact is minimal: just a few milliseconds of added latency. This minor tradeoff provides a major benefit—peace of mind and competitive differentiation.

For more conversations on how security, AI, and cloud intersect, explore more stories on TFiR.