Losses from cybercrime exceeded $20.8 billion in 2025, according to the FBI’s 2025 Internet Crime Report. That’s a 26% jump from the previous year. Meanwhile, research cited by JPMorgan shows that false positives consume around 19% of fraud costs for merchants versus 7% for genuine losses. Classical machine learning is straining under high-velocity payments, cross-border complexity, and privacy constraints due to GDPR and PSD2.
Hybrid quantum–classical approaches are emerging as a promising way forward. Quantum machine learning sharpens sensitivity to faint, non-linear signals in time-series and transaction graphs, with early research reporting faster training and very high classification performance. Leading banks are already experimenting, such as HSBC with Quantinuum – but how close are we to fully quantum-powered fraud detection?
Read on to see our experts explore the potential of quantum computing for the financial services sector and what the future looks like for quantum fraud detection, or find an executive summary on TechInformed here.
Why is the hybrid quantum + AI approach becoming relevant for financial fraud detection?
With 79% of organisations hit by payments fraud, traditional detection systems are buckling under pressure. The surge is driven by instant payments, sophisticated fraud rings, and cross-border e-commerce schemes that evolve faster than rule-based systems can adapt, raising the strategic question of how quantum computing can improve fraud detection for banks facing high-velocity risk.
The false positive crisis compounds the problem. Every wrongly flagged transaction means wasted investigation time, operational bottlenecks, and frustrated customers. For a mid-sized bank processing millions of transactions daily, this means valuable analyst hours spent reviewing low-risk alerts, while genuine fraud cases risk being delayed or overlooked.
In response, institutions are exploring quantum fraud detection in banking and quantum security practices to improve precision at scale. Traditional machine learning has made progress, but it is hitting structural limits, struggling with real-time, high-dimensional data and creating governance challenges under GDPR and PSD2 when models depend on centralised data aggregation.
Fraud teams know the pattern: volumes spike, attackers pivot, models lag and investigators drown in alerts. Upgrading rules and retraining models helps, but the cycle repeats, prompting leaders to ask how to improve fraud detection without breaking privacy rules.
Enter quantum computing. Qubits, the basic units of quantum information, can represent multiple states at once, known as superposition, and become linked to one another, known as quantum entanglement. For certain problems, quantum algorithms running on a quantum processor or quantum chip can optimise and detect patterns more efficiently than classical machines.
Used alongside classical AI, this approach to quantum fraud detection in banking could help banks adapt scenarios faster, cut false positives with smarter triage, and enable compliant cross-border collaboration under GDPR/PSD2.
For a practitioner view on ecosystem readiness, see our conversation preparing for the quantum era with Alice & Bob and Sopra Steria.
3 concrete solutions to today's fraud challenges
We’ve covered the basics for quantum technology’s use in fraud detection, but how can it be used in practice to solve real fraud challenges? Here are three practical ways.
1. Better detection with quantum machine learning
Quantum algorithms excel at spotting subtle patterns in complex datasets, including:
- Time-series anomalies
- Graph structures in transaction networks
- Highly imbalanced fraud signals buried in billions of legitimate transactions
Hybrid classical–quantum models have already demonstrated faster training times and higher accuracy in early research.
Leading institutions are already moving ahead. HSBC partnered with Quantinuum in 2023 to explore quantum computing for fraud detection and cybersecurity. In a 2023 study, quantum machine learning models achieved very high classification performance: an F1-score of 0.98 for both fraud and non-fraud cases – a highly promising result
2. Smarter alert prioritisation through quantum optimisation
Flagging anomalies is only half the battle. The real challenge is deciding which of thousands of daily alerts to investigate first: a combinatorial optimisation problem that grows with complexity. Quantum optimisation within a hybrid quantum fraud detection in banking stack could help efficiently tackle these prioritisation puzzles, helping institutions focus investigation resources where they add the most value.
This directly addresses the false positive crisis: instead of investigating alerts sequentially or through crude scoring, quantum-enhanced systems can weigh multiple risk factors simultaneously to surface the genuinely suspicious cases. These include:
- Transaction history
- Network relationships
- Timing patterns
- Merchant profiles
Coupled with quantum machine learning and aligned with quantum security objectives, the result is fewer wasted hours on false alarms and faster response to real threats.
3. Refresh agility with quantum processing
Beyond accuracy, hybrid quantum approaches can reduce compute time for retraining and re-prioritisation loops by accelerating certain optimisation and anomaly-detection subroutines on a quantum processor.
In practice, this can enable more frequent scenario updates (such as thresholds, graph partitions and alert triage) subject to model governance and integration constraints, helping keep precision high as tactics evolve.
A pragmatic roadmap for financial institutions
Quantum technologies are shifting from theory to practice. Our recommendation is to adopt a phased approach.
1. Identify one pain point, and frame it as a quantum problem
Start small and specific. Focus on a high-impact fraud use case where classical models are reaching their limits, for example:
- Instant payments (high data velocity, tight response windows)
- Card-not-present transactions (imbalanced datasets and pattern drift)
- Cross-border transfers (complex network topologies and data-sharing restrictions)
Each of these domains maps well to known quantum strengths: quantum optimisation, graph analytics, and high-dimensional pattern recognition. Framing the effort as quantum fraud detection in banking clarifies scope and success metrics.
2. Build a hybrid pilot
Work with your existing data pipelines and fraud analytics tools. The objective is to integrate quantum subroutines for anomaly detection, model retraining, or alert prioritisation into your current workflows, showing concretely how quantum computing improves fraud detection.
- Use hybrid cloud access via providers like Quandela to run quantum–classical experiments.
- Measure improvement using clear KPIs: reduction in false positives, time to detect, or analyst productivity.
- Keep the pilot explainable: regulators will want transparency on how alerts are generated and validated.
3. Build the skills and governance now
Quantum experimentation works best when embedded in a clear governance model. Establish a working group to define how quantum machine learning and quantum optimisation results will be validated within your model risk framework.
4. Move from pilot to production-ready insights
Once a pilot demonstrates added value, the next step is industrialisation. The key is incremental adoption: use quantum where it clearly outperforms classical methods and continuously benchmark the ROI with stable KPIs.
Quantum in fraud detection isn’t a distant vision: it’s an R&D-driven performance lever available today.
The window for quantum fraud detection is now
Quantum-enabled fraud detection may not be fully industrialised yet, but it is within our reach – and the time to strike with quantum machine learning pilots is now.
Financial fraud is growing faster than classical detection systems can adapt. Early movers will reduce mounting losses and curb escalating false positive costs by pairing quantum machine learning and quantum optimisation with existing pipelines.
The question isn't whether quantum will transform fraud detection: it's whether your institution will lead that transformation or follow. In parallel with detection, we are strengthening the data protection layer with hybrid post-quantum cryptography.
Start today. Speak to one of our experts.
FAQ
What is quantum fraud detection in banking, and how does it differ from classical machine learning?
It is a hybrid approach that combines quantum machine learning and quantum optimisation with classical ML to spot faint, non-linear patterns in time-series and transaction graphs.
Which technology can be used to detect fraud in financial services?
Hybrid quantum fraud detection in banking using quantum machine learning for pattern recognition and quantum optimisation for alert triage.
How does quantum optimisation prioritise alerts to cut false positives and speed investigations?
It reframes triage as a combinatorial optimisation problem, evaluating multiple risk factors at once to surface the most suspicious cases first, reducing false alarms and investigator workload.
How do we run a hybrid quantum–classical pilot via cloud access with explainable governance suitable for regulators?
Use cloud access (e.g., Quandela) to plug quantum subroutines into current workflows, track hard KPIs (false-positive reduction, time-to-detect, analyst productivity), and document model risk, explainability, and GDPR/PSD2 alignment for scalable governance.