Enigmata raises $6.5M to run AI without exposing plaintext

Why it matters

Enterprises have large stores of regulated and proprietary data that AI teams cannot safely move into ordinary model pipelines. Enigmata's bet is that cryptography can remove that bottleneck without imposing the performance penalties that have limited encrypted computation, but its internal benchmarks and security model still need external proof.

Enigmata co-founder and CEO Scott Searle is building a cryptographic layer for AI systems working with sensitive enterprise data. On September 10th, Enigmata emerged from stealth with a $6.5 million seed round led by Blockchange Ventures and a product designed to let enterprises train models, search documents and analyze records without exposing the underlying plaintext.

The funding will support commercialization of Enigmata Cipher, according to Enigmata's announcement. Enigmata has opened Cipher to selected enterprise design partners, targeting banks, insurers, health systems, life sciences groups, publishers and data providers whose most valuable records are often the least available to outside AI systems. Enigmata has not disclosed customer names, pricing or a general-availability date.

Searle previously founded LockStream, a digital-rights-management and encryption company. Patent records identify him as an inventor on LockStream encryption and content-accountability technology, including a patent involving encryption with personally valuable user keys.

Encryption that does not stop the work

Conventional encryption protects stored or transmitted information, then typically requires an authorized system to decrypt it before meaningful computation begins. That plaintext stage creates the exposure Enigmata wants to remove from AI workflows.

Enigmata says Cipher transforms fields, records, documents and feature sets into protected representations that standard machine-learning, analytics and search systems can still use. Enigmata lists model training, inference, semantic search, retrieval-augmented generation, fraud analysis and third-party data collaboration among the intended workloads. Policy controls govern when an approved user or system can reveal source fields, with each reveal recorded in an audit trail.

The practical pitch is compatibility. Enigmata says Cipher can run on existing enterprise hardware and supports AI, analytics and search workflows across enterprise data environments. Its public materials do not provide enough architectural detail to compare its security assumptions directly with fully homomorphic encryption, secure multiparty computation or trusted execution environments. Enigmata describes support across Databricks, SageMaker, Snowflake, Vertex AI, Azure Machine Learning, on-premises GPUs, private clouds and controlled data centers.

Searle framed Enigmata's thesis in direct terms in the company's announcement: "The world's most valuable data no longer has to stay locked away to stay protected."

That claim carries a demanding technical burden. Enigmata says its internal benchmarks found that models trained on Cipher-protected data matched the accuracy of models trained on raw data while completing training 8% to 10% faster. Those figures come from Enigmata's own testing. Enigmata has not published the workloads, model configurations, datasets or independent security analysis needed to reproduce the comparison.

The claimed speed improvement is particularly consequential. Privacy-preserving computation has historically imposed a performance cost, especially for systems based on fully homomorphic encryption. Duality Technologies markets encrypted analytics and machine-learning workflows using fully homomorphic encryption and secure multiparty computation. Opaque takes a hardware-backed confidential-computing approach, keeping data encrypted in memory and using attestation to verify workloads.

Enigmata describes Cipher as a separate cryptographic transform designed to preserve data utility on standard infrastructure. Design-partner deployments will have to establish where Cipher sits among those approaches and which threat models it can credibly address.

A second product for data that should never return

Cipher is one part of Enigmata's broader product plan. Enigmata also markets Enigmata Anonymizer, a one-way transformation intended for records that should never be restored to their original form. Cipher covers workflows where an authorized reveal may eventually be required; Anonymizer is aimed at uses where reversibility itself would be a liability.

That distinction matters for regulated buyers. A hospital may need to train a model without exposing patient identities, then permit an authorized clinician to view a specific underlying record. Another workflow may require permanent de-identification before data leaves a controlled environment. Enigmata is trying to sell both options through one governance and audit layer rather than forcing customers to assemble separate privacy tools for each use case.

Blockchange backs a data-rights layer

Blockchange's lead investment adds another dimension to the round. Blockchange was founded around blockchain and digital-asset infrastructure, while Enigmata presents Cipher as enterprise AI security rather than a blockchain product. The overlap appears in Enigmata's longer-term plan for data licensing.

Enigmata wants institutions to license protected datasets for model training while retaining control over the original records and setting enforceable usage terms. Publishers, healthcare organizations and financial institutions could theoretically permit computation against valuable data without handing a buyer an unrestricted plaintext copy.

That market remains prospective. Enigmata has not named licensing customers or commercial contracts, and the September 10th announcement did not disclose a valuation or other participants in the seed round. Blockchange is backing the possibility that cryptographic controls can turn data access into a governed transaction, extending the economic logic of digital ownership into enterprise AI.

Searle's LockStream patents covered encryption and content accountability for digital information. The $6.5 million round gives Enigmata's team room to test whether Cipher can provide comparable controls for datasets, model inputs and AI retrieval systems at production speed.