Archer Materials Hits Quantum Fraud Detection Milestone With Benchmarking Success

Archer Materials advances quantum fraud detection with successful benchmarking results

Archer Materials has completed the next stage of its quantum machine learning fraud detection research project, following the dataset preparation milestone announced in March 2026. The quantum technology company successfully tested and benchmarked a quantum neural network model using a publicly available financial fraud dataset containing more than 280,000 transaction records. The early-stage QNN performed equivalently to the best classical models used in benchmarking, generating only one false positive while correctly identifying 118 fraudulent transactions.

The results demonstrate that a QNN model can be used for fraud detection, can achieve high precision, and can run on both quantum simulators and real quantum hardware. The company missed 30 fraudulent transactions in the simulator test environment.

Why false positives matter in fraud detection

False positives — legitimate transactions incorrectly flagged as fraudulent — create significant operational and customer experience costs for financial institutions. Each false alert requires manual review, consuming staff resources and increasing operational expenses. Excessive false positives also damage customer experience when legitimate purchases are declined or accounts are temporarily frozen.

The QNN result addresses this industry pain point directly. Detecting 118 fraudulent transactions while generating only one false positive demonstrates high precision in a real-world dataset. This balance between fraud capture and false alert reduction is a key differentiator for any fraud detection solution.

Financial institutions investing in fraud detection systems must weigh two competing risks: missed fraud versus unnecessary intervention. The early QNN performance suggests the quantum approach may help optimise this trade-off, a commercially valuable outcome for potential partners in banking and payments.

Quantum hardware validation on commercial systems

The model was successfully executed on IQM Garnet, a 20-qubit superconducting quantum computer accessed through AWS Braket. The hardware test detected 18 of 19 fraudulent transactions in the test set, validating that the model operates on commercial quantum hardware, not just simulators.

The real-hardware experiment involved higher false-positive rates than simulator testing, acknowledged as a limitation by the company. This outcome reflects current quantum computing constraints rather than a fundamental model weakness. Demonstrating execution on physical quantum systems — not just theoretical simulators — is a critical step toward eventual commercial deployment.

Noise tolerance findings

Quantum noise is a key challenge for practical QML applications, arising from imperfections in quantum hardware that introduce errors into calculations. The model remained stable at low noise with minor performance degradation at moderate noise levels. Performance declined materially at higher noise levels, a useful technical finding that helps identify hardware quality requirements for future practical applications.

Understanding noise tolerance thresholds informs hardware selection and deployment planning for commercial use cases. The staged experimental framework used qubit-selection studies, feature-map optimisation, benchmarking against classical machine learning approaches, and quantum noise analysis to identify optimal model configurations.

Global financial institutions exploring similar applications

Archer’s work sits within a broader industry trend of quantum machine learning exploration by leading financial institutions:

• Quantinuum’s collaboration with HSBC to explore QML for fraud detection
• Intesa Sanpaolo’s collaboration with IBM to explore QML for improving fraud detection accuracy and speed

The company is pursuing a use case that major global banks are actively investigating, validating market interest and potential commercial pathways. While Archer has not yet demonstrated a clear performance advantage over leading classical AI approaches, the company has substantially reduced technical uncertainty.

Technical approach and benchmarking framework

The dataset was processed using dimensionality reduction and data balancing techniques to operate within current quantum computing constraints. The staged experimental framework established a repeatable benchmarking approach that can be applied to larger systems and more complex datasets.

The dataset preparation milestone completed in March 2026 applied dimensionality reduction and data balancing techniques specifically to address qubit-count constraints on present-day quantum hardware, a preprocessing step that shaped the benchmarking framework used in the current phase.

The work identified a high-performing quantum architecture, established a repeatable benchmarking framework, and provided insights into scaling, noise tolerance, and deployment constraints. The company has successfully demonstrated that QML models can perform competitively on a real fraud-detection benchmark and can be executed on both simulators and physical quantum hardware.

CEO outlines path toward full QML prototype

Archer is aiming for a full QML prototype by the end of 2026.

Dr Simon Ruffell, CEO of Archer

“These results have demonstrated that QML approaches can deliver strong fraud detection performance while operating within the constraints of current quantum computing systems.

“The simulator results were solid, particularly the very low false-positive rate, and the successful execution on real quantum hardware is an important validation step.

“The research collaboration agreement with the CSIRO forms part of Archer’s strategy to investigate practical applications of quantum computing technologies and support future commercialisation opportunities in data-intensive industries. Fraud detection is a relevant use case for QML because banks and payment providers must analyse large volumes of transaction data quickly, while reducing both missed fraud and false alerts.”

The company has not yet demonstrated a clear performance advantage over leading classical AI approaches, but has substantially reduced technical uncertainty and established a foundation for the next phase of the project.

Next steps and development pathway

Further testing is required before any commercial deployment pathway can be assessed. The next development steps include:

1. Testing on larger datasets to validate scalability
2. Additional classical benchmarks to strengthen comparative analysis
3. Repeated trials to confirm performance consistency
4. Further hardware validation on commercial quantum systems

This phase was conducted on a prepared research dataset, using a selected group of comparison models and current quantum computing constraints. The foundation phase will test whether larger quantum systems and richer feature representations can deliver measurable business advantages over classical approaches.

Archer’s quantum prototype milestones span multiple parallel workstreams beyond fraud detection, with the 12CQ qubit platform, quantum sensing applications, and room-temperature operation capabilities each progressing under a distributed partnership model designed to reduce capital requirements.

Ready to Learn More About Archer’s Quantum Computing Roadmap?

Archer Materials has successfully demonstrated that quantum machine learning can compete with classical fraud detection systems whilst running on commercial quantum hardware. The company is now progressing toward a full QML prototype by the end of 2026.

To explore Archer’s broader quantum technology portfolio—including 12CQ qubit development, quantum sensing applications, and partnership strategy—visit the Archer Materials investor centre for company updates, technical milestones, and commercialisation timelines.