Conference Description
Key Takeaways
- AMLUCS is the UK’s only government-backed conference dedicated to the intersection of artificial intelligence and cybersecurity
- The fourth annual event focuses on practical, real-world applications rather than theoretical or commercial presentations
- Core themes include AI security threats, offensive and defensive AI operations, multi-agent system architectures, governance frameworks and explainability
- Aimed at security engineers, data scientists, researchers, policymakers and operational practitioners
- Features a pre-conference training course on Validated AI Red Teaming for Production AI Systems
Introduction
AMLUCS returns for its fourth year as the United Kingdom’s sole government-backed conference addressing the convergence of artificial intelligence and cybersecurity. The event serves security engineers, data scientists, researchers, vendors, policymakers and operational practitioners who work directly with AI-enabled systems in defensive and offensive contexts. As organisations increasingly deploy machine learning models across critical infrastructure and security operations, the need for rigorous examination of both the opportunities and vulnerabilities these systems present has become acute. AMLUCS positions itself at this intersection, prioritising deployment experience and applied research over commercial messaging.
About AMLUCS
Operating as a not-for-profit, practitioner-led initiative, AMLUCS distinguishes itself through its emphasis on operational reality. The conference is supported by an international research community and guided by a Technical Committee responsible for programme curation. This structure ensures that content reflects genuine challenges encountered by professionals building, testing and defending AI systems rather than vendor-driven narratives.
The event functions as a bridge between applied research and operational practice, recognising that academic advances in machine learning security often require significant adaptation before they prove useful in production environments. By facilitating direct exchange between researchers and practitioners, AMLUCS aims to accelerate the translation of theoretical insights into deployable defences and assessment methodologies.
Conference Themes and Technical Focus
The programme is organised around five interconnected themes that reflect the current state of AI security practice and emerging concerns within the field.
AI Security Threats, Vulnerabilities and Mitigations
Machine learning models introduce novel attack surfaces that traditional security frameworks were not designed to address. Adversarial examples, data poisoning, model extraction and membership inference attacks represent categories of threat that require specialised detection and mitigation strategies. This track examines both the theoretical underpinnings of these vulnerabilities and the practical countermeasures that have proven effective in operational settings.
Agent Capabilities and Offensive/Defensive AI Operations
The deployment of AI agents in cyber operations—whether for automated threat hunting, vulnerability discovery or adversarial simulation—raises significant questions about capability assessment and operational control. As autonomous systems take on more complex tasks within security workflows, understanding their behavioural boundaries and failure modes becomes essential for both offensive and defensive applications.
Multi-Agent Systems and Secure Architecture Design
Increasingly, AI deployments involve multiple interacting agents rather than isolated models. These architectures introduce coordination challenges, trust boundaries and potential cascade failures that single-model security assessments fail to capture. Secure design principles for multi-agent systems remain an active area of research with direct implications for enterprise deployments.
Governance, Risk Assessment and Emerging Standards
Regulatory frameworks governing AI systems are evolving rapidly across jurisdictions, with particular attention to high-risk applications in security contexts. Organisations deploying AI within their security operations must navigate compliance requirements while maintaining operational effectiveness. This track addresses risk assessment methodologies, governance structures and the emerging standards landscape that shapes responsible AI deployment.
Explainability, Interpretability and Fairness
Security decisions informed by AI systems often require justification, whether for incident response documentation, regulatory compliance or operational trust. The ability to explain model behaviour and ensure fair treatment across different inputs carries particular weight in security applications where false positives and negatives carry significant consequences. This theme explores techniques for achieving meaningful transparency without compromising model performance.
Programme Structure
The conference programme combines multiple formats designed to address different aspects of professional development and knowledge exchange. Technical talks draw on applied research and deployment experience, presenting findings grounded in real-world implementation rather than laboratory conditions alone. Keynote sessions feature international experts alongside UK government leaders, providing perspective on both technical frontiers and policy direction.
Practitioner panels offer opportunities to examine current challenges and emerging risks through facilitated discussion among professionals actively engaged with these issues. The format allows for exploration of nuanced topics where consensus has not yet formed and where operational experience provides essential context that formal research may not capture.
Pre-Conference Training
A two-day training course on Validated AI Red Teaming for Production AI Systems precedes the main conference. Red teaming methodologies adapted for AI systems represent a growing discipline as organisations seek to identify vulnerabilities before adversaries exploit them. The course addresses the specific challenges of assessing production AI systems, where constraints around access, documentation and operational continuity differ substantially from research environments.
Who Should Attend
AMLUCS is designed for professionals who work directly with AI systems in security contexts. Security engineers responsible for protecting AI-enabled infrastructure will find relevant content on threat modelling and defensive techniques. Data scientists building models for security applications can engage with research on robustness and adversarial resilience. Researchers benefit from exposure to operational constraints that shape how their work translates into practice.
Policymakers and governance professionals gain insight into technical realities that inform effective regulation and standards development. Vendors developing AI security products encounter the practitioner perspective essential for building tools that address genuine operational needs. The common thread is active engagement with the challenges of building, assessing or defending AI systems rather than passive interest in the field.
The Broader Context
The intersection of AI and cybersecurity has moved from speculative concern to operational reality. Organisations across sectors now deploy machine learning for threat detection, anomaly identification and automated response. Simultaneously, adversaries leverage AI capabilities to enhance attack sophistication and scale. This dual-use nature demands that security professionals understand both the defensive potential and the novel risks these technologies introduce.
Government backing for AMLUCS reflects recognition that national cybersecurity posture increasingly depends on the secure development and deployment of AI systems. The conference contributes to building the expertise and collaborative networks necessary to address challenges that no single organisation can solve in isolation.