Tue Aug 04 2026
The uncomfortable truth about today’s enterprise AI is that it was never built to keep secrets — and now there’s an independent research program designed to prove exactly how much it gives away. 01 Quantum Inc. (TSXV: ONE; OTCQB: OONEF), Carleton University, and the National Centre for Critical Infrastructure Protection, Security and Resilience (NC-CIPSeR) have announced a strategic research collaboration to stress-test defensive technologies against a problem most organizations haven’t priced in: modern AI systems that quietly reveal the sensitive data they were trusted to protect. The work is designed to quantify that leakage — and to demonstrate how Fully Homomorphic Encryption (FHE) and Post-Quantum Cryptography (PQC) can close the gap.
Most enterprise AI deployments share a hidden design flaw. When multiple AI agents collaborate on a task — fraud detection, customer risk scoring, claims triage — they routinely exchange sensitive customer data in plain text. Every hand-off between agents is a moment of exposure, and every exposure is an opportunity for an attacker. The Carleton and NC-CIPSeR project will simulate a realistic financial institution environment with exactly this kind of multi-agent workflow, then measure what falls out.
The specific vulnerability under the microscope is membership inference — where an outside party can statistically determine whether a particular customer’s record was part of a model’s training data. For a bank, a government agency, or a defense contractor, that’s not an abstract privacy nicety. It’s a direct line from a deployed AI model back to an individual, and it’s the kind of leakage that regulators and adversaries alike are learning to exploit.
What makes this collaboration different is who’s doing the measuring. The first phase quantifies data leakage from a standard, non-encrypted AI system — establishing the baseline damage. A later phase tests an encryption-compatible model component to demonstrate how FHE can preserve privacy while the model still does useful work.
That sequencing matters. Rather than asserting that encrypted AI is safer, the project is structured to produce academically validated, third-party evidence of both the vulnerability and the remedy. For enterprises evaluating whether to trust their most sensitive workloads to AI, independent research from a leading Canadian institution carries a weight that vendor whitepapers never will.
The reason 01 Quantum’s architecture is relevant here comes down to a gap most security stacks ignore. Encryption at rest protects stored data. Encryption in transit protects data moving across networks. But AI systems have to actually process data — and traditionally, that means decrypting it first, exposing it precisely when it’s most valuable. Fully Homomorphic Encryption changes the equation by allowing computation directly on encrypted data, so the information stays protected even while the model is working with it. That’s the “data-in-use” layer that plaintext AI leaves wide open.
01 Quantum’s approach pairs FHE with its IronCAP™ PQC technology — patent-protected in the U.S.A. under #11,271,715 and #11,669,833 and aligned with NIST’s FIPS 203, 204, and 205 standards — to bring AI workloads into a genuinely quantum-safe operating environment. It’s the same foundation that underpins the Company’s work across digital assets, but the AI privacy application represents the larger addressable market: every institution deploying multi-agent AI on sensitive data is a candidate.
The timing lands alongside a shifting policy landscape. On June 22, 2026, an Executive Order directed U.S. federal agencies to accelerate preparations for quantum-enabled cybersecurity threats — instructing them to identify systems relying on vulnerable encryption, prioritize migration to quantum-resistant standards, and coordinate with industry partners to keep operations secure. The Order also flagged the “harvest now, decrypt later” risk, where adversaries collect encrypted data today to break it once quantum capabilities mature.
That’s the same threat model 01 Quantum has been building against, and the same NIST-aligned standards its technology is designed to fulfill. As CEO Andrew Cheung framed it, encrypted AI should become essential infrastructure in the years ahead — and this partnership is intended to supply the independent evidence that proves the point.
The through-line across this collaboration is that 01 Quantum isn’t forming a committee or promising a roadmap — it’s bringing working PQC and FHE technology to an independent research program designed to validate it under realistic conditions. While much of the industry is still debating whether AI privacy is a real problem, this project is built to measure the damage and demonstrate the fix in the same breath.
The AI systems now running inside banks, agencies, and defense environments were built to be useful, not discreet — and the organizations that recognize the difference first will be the ones still trusted when the leakage becomes undeniable.