Enigmata has emerged from stealth with $6.5 million in seed funding led by Blockchange Ventures to commercialize Enigmata Cipher, a patent-pending cryptographic technology designed to let AI systems train, search and analyze data while it remains encrypted. The approach targets a growing enterprise problem: making sensitive datasets available to AI without requiring organizations to expose the underlying information to models, applications or infrastructure operators.

Enterprise AI has created a difficult security paradox: the data that could make AI most valuable is often the data organizations are least willing to expose.

Healthcare records can support clinical research and disease detection. Financial transaction data can improve fraud detection. Proprietary documents can provide the knowledge required for enterprise AI systems and agents. But putting those datasets into conventional AI pipelines can create new privacy, security, regulatory and intellectual-property risks.

Nashville-based Enigmata is emerging from stealth with a technology designed to address that problem.

The company has raised $6.5 million in seed funding led by Blockchange Ventures to commercialize Enigmata Cipher, a patent-pending cryptographic system that Enigmata says enables artificial intelligence to train, search and analyze information while the underlying data remains encrypted.

The technology is aimed at workflows including proprietary-data model training, enterprise use of third-party AI systems, semantic search, analytics and agentic applications.

The fundamental idea is part of a broader field known as privacy-enhancing cryptography.

Traditional encryption protects data while it is stored or transmitted, but data generally has to be decrypted before conventional software can process it. Privacy-enhancing technologies attempt to extend protection into computation itself.

The National Institute of Standards and Technology (NIST) identifies fully homomorphic encryption, secure multiparty computation and other cryptographic techniques as tools for computing over protected information. NIST specifically describes fully homomorphic encryption as allowing computation on encrypted data without exposing the underlying plaintext.

Enigmata is positioning Cipher as an AI-oriented implementation of that broader concept.

According to the company, Cipher transforms records, documents and datasets into an encrypted representation that AI, analytics and search systems can use on existing enterprise hardware. Enigmata says the technology simultaneously structures and encrypts data, which it claims can improve performance for AI workloads.

The company’s internal benchmarks reportedly found that models trained using Cipher-protected data achieved accuracy comparable to models trained on raw data while completing training 8% to 10% faster.

Those performance figures are Enigmata’s own benchmarks rather than independently validated results, an important distinction for enterprises evaluating a new cryptographic architecture.

The larger technical challenge is significant. Encrypted computation has historically involved trade-offs involving computational overhead, supported operations, flexibility and deployment complexity.

Recent research illustrates both the potential and the remaining challenges. A 2026 study of homomorphic-encryption-based machine learning reported comparable model performance while identifying computational overhead, noise management and limitations around certain operations as continuing obstacles.

Another peer-reviewed study published in Scientific Reports this month evaluated encrypted machine-learning inference using homomorphic encryption and highlighted the growing effort to make encrypted AI practical in cloud environments.

That makes Enigmata’s claimed ability to operate on existing enterprise hardware an important part of its positioning.

If organizations can use encrypted data without rebuilding their entire AI infrastructure around specialized systems, privacy-preserving computation could become more accessible for enterprises deploying third-party models and AI agents.

The use case extends beyond model training.

Semantic search is particularly relevant because enterprise retrieval-augmented generation systems increasingly ingest internal documents, customer information and proprietary knowledge. A conventional RAG architecture can create multiple locations where sensitive information is processed, cached or exposed.

Encrypted search could potentially reduce the amount of plaintext information that needs to be exposed during those operations.

The same principle could apply to AI agents. As agents gain access to CRM, ERP, financial, healthcare and other enterprise systems, protecting the data they retrieve becomes as important as controlling what actions they can perform.

This creates an emerging architecture in which identity, authorization, encrypted computation and AI governance become interconnected.

Enigmata is also proposing a longer-term commercial model around data licensing.

The company’s vision is for institutions to make encrypted datasets available for AI training under enforceable usage conditions while retaining control over the underlying assets. That could become significant as copyright disputes and negotiations over AI training data increasingly affect publishers, content owners and model developers.

Instead of permanently transferring datasets to AI companies, organizations could potentially provide controlled computational access to protected data.

That would change the economics of enterprise data.

Today, an organization often faces a binary choice: keep valuable information isolated or provide a model provider with some form of access. Privacy-preserving computation could create a third option in which data remains under the owner’s control while still generating value through AI.

The concept is consistent with NIST’s broader view of privacy-enhancing cryptography. NIST identifies applications including privacy-preserving AI, medical collaboration and financial fraud detection, while noting that different privacy technologies have different security and performance characteristics.

Confidential computing represents another competing approach. Instead of performing computation directly over ciphertext, trusted execution environments protect data while it is processed inside hardware-based secure enclaves. NIST’s 2026 draft guidance specifically examines confidential computing for protecting AI workloads and sensitive datasets in cloud environments.

That distinction matters for enterprises evaluating Enigmata.

Cipher is entering a market that already includes fully homomorphic encryption, confidential computing, secure multiparty computation, federated learning, differential privacy and other privacy-enhancing technologies. These approaches are not interchangeable, and the right architecture depends on the threat model, workload, performance requirements and level of trust an organization is willing to place in infrastructure providers.

Enigmata’s opportunity is therefore less about replacing every privacy technology than about making encrypted computation practical for mainstream AI workloads.

The company’s initial funding provides resources to turn its cryptographic technology into a commercial platform. But enterprise adoption will ultimately depend on independent security evaluation, reproducible performance benchmarks, interoperability with existing AI stacks and evidence that encrypted workloads can operate at production scale.

If those requirements can be met, the implications extend well beyond privacy.

Encrypted AI could allow organizations to treat sensitive information as an AI-accessible asset without treating it as an AI-exposed asset.

That could give hospitals, banks, publishers and other data-intensive organizations a new way to participate in the AI economy while maintaining tighter control over some of their most valuable information.

Market Landscape

The market for privacy-preserving AI is expanding across several technical approaches.

Fully homomorphic encryption (FHE) enables computation directly on encrypted data and is particularly relevant when organizations do not want an infrastructure provider to see plaintext information. NIST identifies FHE as a major privacy-enhancing cryptography technique with applications in AI, healthcare and fraud detection.

Confidential computing takes a different approach by protecting data while it is being processed inside trusted hardware environments. NIST is actively developing guidance for confidential computing in cloud workloads, including AI applications.

Federated learning keeps training data distributed rather than centralizing it, while differential privacy adds mathematically quantifiable privacy protections to datasets and model outputs.

Enigmata is positioning Cipher around encrypted computation specifically for AI, search and analytics.

The competitive question will be whether its architecture can deliver a sufficiently attractive combination of security, flexibility, model compatibility and performance without forcing enterprises to redesign their existing AI infrastructure.

Top Insights

• Encrypted AI targets a major enterprise barrier: Organizations can gain value from sensitive datasets without necessarily exposing plaintext data to AI infrastructure.
• Privacy-enhancing computation is broadening: FHE, confidential computing, federated learning and differential privacy provide different approaches to protecting AI data and workloads.
• Performance remains critical: Enigmata claims Cipher-trained models can match raw-data accuracy while completing training 8% to 10% faster, but the results are company benchmarks.
• AI agents increase the stakes: As autonomous systems gain access to sensitive enterprise data, protecting information during computation becomes an important part of agent security.
• Data licensing could become a new market: Encrypted computation could eventually allow data owners to provide controlled AI-training access without surrendering underlying datasets.

Power Tomorrow’s Intelligence — Build It with TechEdgeAI