01 Quantum Secures U.S. Patent for Encrypted AI Inference Technology

01 Quantum has received U.S. Patent No. 12,744,764 for a system designed to perform AI inference while keeping both user data and AI models encrypted during processing. The patented architecture combines computation-preserving cryptography with post-quantum cryptographic communication, with potential applications in privacy-sensitive AI environments.

AI Security Moves Beyond Protecting Data in Transit

As organizations increasingly use AI to process medical records, financial information, industrial data and other sensitive material, encryption is becoming a more complicated problem. Protecting information while it travels between systems is only one part of the challenge. The data may still need to be exposed to an AI system during computation.

01 Quantum Inc., a Toronto-based cybersecurity company, is addressing that problem with a patented architecture designed to keep sensitive information encrypted during AI inference.

The United States Patent and Trademark Office has issued U.S. Patent No. 12,744,764 B1, titled “Secure AI Operations.” The patent was issued on September 22, 2026, to 01 Quantum, with Andrew Cheung and Alexey Shpurov listed as inventors.

The company announced the patent award on September 29.

How the Patented System Works

The central concept is computation-preserving cryptography (CPC). Unlike conventional encryption, where information generally has to be decrypted before a computation can be performed, CPC is designed to allow mathematical operations to take place on encrypted information.

In the architecture described by the patent, a user’s sensitive data is encrypted with a CPC public key before being submitted for an AI prediction. The AI model can also be encrypted using the same cryptographic framework.

The encrypted model then processes the encrypted input, producing an encrypted prediction. The user can subsequently decrypt the result using the corresponding private key.

In practical terms, the objective is to reduce the amount of sensitive information exposed during an AI inference workflow.

The patent also describes communication between the participants using post-quantum cryptographic (PQC) secure sessions, adding a separate layer intended to protect communications against future quantum-computing-based attacks.

Protecting the AI Model Itself

The technology is not focused only on protecting user information.

AI models can themselves represent valuable intellectual property. Companies may be reluctant to make proprietary models available to external infrastructure if doing so could expose model parameters or other sensitive components.

The patented architecture therefore covers scenarios in which an AI model supplied by a model vendor is encrypted before being executed against encrypted user data.

The patent also describes an AI model marketplace in which model vendors could make models available to users while maintaining cryptographic protection over both the models and user information.

That creates a potential architecture in which the party providing the AI model does not necessarily need direct access to the underlying customer data, while the customer does not necessarily need access to the model’s protected internals.

Why Post-Quantum Cryptography Matters

The patent combines two related but different security concepts.

Computation-preserving cryptography addresses the privacy of information during computation. Post-quantum cryptography, meanwhile, is intended to protect cryptographic communications against attacks that could become practical with sufficiently capable quantum computers.

01 Quantum’s existing technology portfolio includes its IronCAP platform, which the company describes as a post-quantum cryptography system incorporating NIST-standardized algorithms and its own patented technologies.

The distinction is important because post-quantum encryption by itself does not automatically mean that AI inference can be performed on encrypted data. The patent combines the two approaches: CPC for encrypted computation and PQC for securing communication sessions around the operation.

From Medical AI to Financial Systems

The potential applications are concentrated in areas where AI needs access to sensitive information.

Healthcare is one example. An AI system could potentially analyze protected clinical information without requiring the raw data to be exposed in the same way as a conventional processing workflow.

Financial institutions could similarly use privacy-preserving AI for applications involving transaction data, fraud detection or other confidential information.

Other possible areas include government systems, defence-related applications, industrial analytics and proprietary enterprise AI.

01 Quantum has previously described its Quantum AI Wrapper technology as being aimed at special-purpose AI applications where data sensitivity and model protection are particularly important. Its examples include fraud detection and medical diagnostics.

However, these remain technology applications and commercial possibilities rather than evidence that every such use case has already been deployed at scale.

The Technology Still Has a Performance Challenge

Encrypted computation comes with an important trade-off: processing encrypted information can be significantly slower than performing the same computation on unencrypted data.

The patent itself acknowledges this limitation and notes that, at the time of its filing, CPC’s computational overhead constrained its use particularly in large language models. The document instead points toward special-purpose AI systems as a more practical target.

That limitation matters because the most demanding AI systems today can involve enormous models and substantial computational workloads.

The commercial value of encrypted AI will therefore depend not only on security, but also on whether the underlying cryptographic operations can be made fast enough for practical workloads.

A Patent Is Not the Same as Commercial Deployment

The issuance of the patent establishes intellectual-property protection for the described system and methods. It does not, by itself, demonstrate that the technology has achieved a particular level of performance, security, adoption or commercial scale.

01 Quantum says its Quantum AI Wrapper technology is being developed for secure AI operations and has reported independent validation work involving Carleton University.

The company has also positioned the technology within a broader portfolio covering quantum-safe remote access, digital assets, email security and AI.

The newly issued patent strengthens that portfolio by specifically covering an architecture for encrypted AI operations.

The Bigger Shift Toward Private AI

The development reflects a broader challenge emerging alongside the rapid adoption of AI: organizations increasingly want to use sophisticated models without surrendering control over the data and intellectual property involved.

Traditional security approaches largely focus on protecting information before and after computation. Privacy-preserving computation attempts to extend that protection into the processing stage itself.

For 01 Quantum, Patent No. 12,744,764 represents a formal step toward that model of AI security. Whether encrypted inference can achieve the speed, scalability and flexibility required for widespread enterprise deployment will remain an important engineering and commercial question.

FAQs

What patent did 01 Quantum receive?

01 Quantum received U.S. Patent No. 12,744,764 B1, titled “Secure AI Operations,” issued by the USPTO on September 22, 2026.

What does 01 Quantum’s AI patent protect?

The patent covers systems and methods for performing AI operations using computation-preserving cryptography, including processing encrypted user data with an encrypted AI model and returning an encrypted prediction.

What is computation-preserving cryptography?

Computation-preserving cryptography is a form of encryption designed to allow computations to be performed on encrypted information without first exposing the underlying data in plaintext.

Is 01 Quantum’s technology the same as post-quantum cryptography?

No. The patent combines computation-preserving cryptography for encrypted computation with post-quantum cryptography for securing communications. They address different parts of the security problem.

Can encrypted AI inference be used with large language models?

Potentially, but performance is an important limitation. The patent notes that the computational overhead of CPC can restrict its use, particularly for large language models, making specialized AI applications a more practical target in the technology’s current context.