Agenda Day 1

AI is reshaping where cyber responsibility begins and ends. Attacks now target prompts, models, and data flows, areas that sit outside traditional controls. As organisations move to secure AI, leaders face real trade-offs between risk reduction and speed. This opening panel surfaces where to draw the line, who owns it, and what's practical today.

Alex Gomez

Divisional CISO & Global Head of IT Security, Risk & Compliance
The Adecco Group

See Bio

Lalit Kumar

Group CISO and Principal Security Architect
Aviva

See Bio

In this session, Alister Shepherd, CISO at the FCA, shares a practical case study on moving AI security from frameworks and standards into live implementation. Drawing on the FCA's experience, he will explore how security teams can monitor, validate and strengthen AI systems once they are in production. He examines why reliability is a core security concern, not just a technical performance issue, and how failures in live models can create unseen risk if they are not identified early. He also reflects on the journey so far, including lessons learned, early challenges and how his team has adapted its approach by evolving the tooling and controls needed to secure AI systems in practice.

Alister Shepherd

CISO
FCA

See Bio

AI dramatically expands the attack surface, yet the skills required to test these systems remain scarce. Enterprises are under pressure to build internal capability, but few understand what AI red teaming requires in practice. This debate challenges assumptions around in house readiness and exposes where maturing may be overestimated. Panellists will discuss the feasibility of building internal AI red teams, the tooling and expertise genuinely required, and where external specialists remain critical.

Vijay Mitra

Head of AI and Digital Risk
Nationwide Building Society

See Bio

Luigi Renna

Director of Security Architecture
Barclays

See Bio

Model Context Protocol (MCP) is rapidly becoming the backbone of agentic AI, connecting models to tools, memory, plugins, and live data. But it introduces a critical security failure: it collapses trust boundaries. Trusted system instructions, user input, and external data all merge into a single prompt the model cannot reliably interpret or defend. The result is a new, highly exploitable layer, where prompt injection, tool spoofing, and data leakage happen inside the model's execution flow, beyond the reach of traditional controls. In this session, Daniyal Naeem, Principal Security Authority-AI, at BT Group shows how to restore trust boundaries in MCP-based agent systems: separating trusted and untrusted context, validating tools and data before model exposure, and enforcing identity, permissions, isolation, and audit controls across every MCP flow.

Daniyal Naeem

Principal Security Authority - AI
BT Group

See Bio

AI is no longer confined to central platforms. Teams can now build, connect, and deploy AI capabilities directly into workflows. Tools like Claude and Microsoft Copilot are just the starting point; open frameworks and APIs are putting real power in the hands of employees. The opportunity is speed and innovation. The risk is uncontrolled sprawl: where AI systems are built, integrated, and used without consistent oversight. This session focuses on how to enable widespread AI adoption while maintaining control over data, usage, and risk.

Gunjan Gandhi

CTIO
Ericsson

See Bio

Join relaxed, topic-led roundtables designed to connect you with peers facing similar challenges in securing AI across the enterprise. In a no-slide, no-stage setting, each table will tackle a timely Secure AI theme. Lightly moderated to spark honest discussion, challenge assumptions, and share what's really happening on the ground.

As AI scales across automotive products, the challenge is no longer building models but proving that they are secure across the entire product lifecycle. In this case study, Sheikh Mahbub Habib, Head of Product Cybersecurity and Privacy Innovation at AUMOVIO, outlines the threats and attacks in the field of AI and how those can affect the automotive products in the long run. He shows how systematically AI threats and attacks can be addressed and mitigated. The session highlights a critical industry challenge: rigorously end-to-end testing AI for safety-critical systems while meeting performance and delivery pressures, and ensuring guardrails are continuously validated to hold up in real-world conditions.

Sheikh Mahbub Habib

Head of Product Cybersecurity and Privacy Innovation
AUMOVIO

See Bio

Enterprise AI is already in production, but security models haven't caught up. Teams are deploying copilots, agents, and AI-powered workflows faster than organisations can define acceptable risk, leading to inconsistent decisions and growing exposure. The challenge isn't understanding every AI threat, it's deciding where to draw the line and enforcing it consistently. This panel focuses on a critical question: how do organisations define and apply AI risk boundaries in practice, across data use, model behaviour, and autonomy, without slowing down delivery?

Vijay Mitra

Head of AI and Digital Risk
Nationwide Building Society

See Bio

Alister Shepherd

CISO
FCA

See Bio