Professor Victor Chang

‘For me, ethics is not a constraint on innovation but a design discipline that produces stronger, more auditable systems.’

Part of the AI Leadership Index – a curated list of the most inspirational figures shaping the AI revolution.

Professor Victor Chang is a Professor of Business Analytics at Aston Business School, where he leads strategic initiatives in applied AI, data analytics and cybersecurity. His career has been built around turning complex AI research into practical systems that protect citizens, patients and businesses.

He has secured £3mn funding as the PI, with active partnerships over 100 institutions worldwide. Internationally recognised as an authority in applied AI, data science, cybersecurity and cloud computing, he has led multi-million-pound collaborations across the UK, Europe, Asia and the US.

His work spans AI-driven malware detection for 5G/6G networks, intelligent IoMT platforms for e-diagnosis, and advanced decision-support systems in finance/risk management. Victor has authored over 300 peer-reviewed publications and delivered over 60 invited keynotes worldwide.

His applied AI solutions have achieved over 90% accuracy in disease diagnosis and up to 99% in malware detection, demonstrating a rare combination of technical depth and real-world impact.

He has over 12 Fellowships from prestigious learned societies. He was named Data Leader of the Year 2025 (In-House) at the British Data Awards and Inspirational Individual of the Year 2024 at the BCS /UK IT Industry Awards, recognising his leadership and contributions in responsible AI and data-driven transformation.

Please share a case study of how you have you innovated with AI over the last year?

Over the past year, I have led a UK–Japan research and innovation partnership developing AI-driven malware detection for 5G, 6G and the Mobile Internet of Things (MIoT), progressing from concept to working prototypes with telecom and cybersecurity partners. Traditional, centralised security operations cannot keep pace with the scale and complexity of attacks in dense 5G/6G and MIoT environments, where bandwidth limits, noisy logs and privacy constraints hinder effective defence.

We addressed this by designing a federated, neuromorphic AI system deployed at the network edge (e.g., O-RAN base stations and MIoT gateways). Each site trains a local model and shares only model updates. Our FedAvgVChang algorithm accelerates federated training by 38% over standard FedAvg, achieving 96.8% detection accuracy (F1: 0.968; AUC: 0.974; precision: 0.966; recall: 0.970).

The system delivers higher-quality alerts with sub-0.8-second median inference times, reducing analyst workload. Supported by an £80,000 International Science Partnerships Fund grant and over £100K in related funding, the project has led to major awards and invited keynotes on federated O-RAN security.

What is the biggest challenge you have faced with AI innovation and how you have overcome it?

My biggest challenge was building trust in high-stakes AI across diverse stakeholders -CISOs, engineers, compliance teams and academics. In the malware detection project, partners welcomed strong performance but were concerned about data sharing, regulation and “black-box” decision-making. Some pushed for full automation; others resisted automated actions in live networks. AI hype also fuelled unrealistic expectations.

I reframed the initiative from “deploying AI” to co-designing secure workflows. We formed a joint steering group spanning security, legal, risk and engineering to agree on focused use cases and defined risk appetite. Governance was embedded into the architecture: federated learning kept sensitive data local, synthetic datasets supported cross-border testing, and model cards, audit logs and kill switches ensured human oversight.

We adopted phased delivery - simulation, sandbox, limited live trial - measuring not only accuracy and latency but also false positives, analyst workload and explainability. Clear communication through workshops, demos and tailored briefings helped stakeholders see operational value. By prioritising governance and culture alongside technology, we turned scepticism into active support and scaled the programme.

How do you balance innovation with ethical safeguards?

For me, ethics is not a constraint on innovation but a design discipline that produces stronger, more auditable systems. I apply a four-layer approach:

• Principles and Scope: Each project starts with a clear statement of purpose, benefits and unacceptable risks, aligned to frameworks such as the EU AI Act and OECD AI principles.
• Governance and Process: We define accountability—who approves use cases, controls data access and can shut systems down—and conduct focused impact assessments on accuracy, privacy, fairness and security.
• Technical Safeguards: In our 6G security work, federated learning and on-device processing keep raw data local. We log decisions, test for bias and drift, and red-team models in controlled settings.
• Culture and Skills: We train end users to understand both AI’s capabilities and limits, encouraging constructive challenge.

This integrated approach delivers high-performance AI while keeping transparency and human judgement at the core.

How do you believe the artificial intelligence space will change over the next 12-24 months?

Over the next two years, AI will shift from isolated pilots to embedded infrastructure, driven by three changes:

• Standardisation and orchestration: Organisations will move from one-off experiments to standardised AI platforms with shared governance and monitoring. Generative AI will evolve from standalone chatbots into integrated “co-pilots” that quietly draft content, refactor code and coordinate workflows inside enterprise systems.

• Architectural split: A small number of large cloud foundation models will coexist with smaller, domain-specific and on-device models that are cheaper, more efficient and easier to govern. Federated learning and privacy-preserving methods will become mainstream to balance performance with data protection.

• Regulation and risk: Frameworks such as the EU AI Act will elevate AI risk management to the level of financial and cyber controls. At the same time, AI-driven threats such as prompt injection, and model theft will grow, alongside AI-enabled defences.

Competitive advantage will depend less on having the largest model and more on combining innovation with strong, auditable governance.