Every talk is grounded in real research and three decades of hands-on experience.

Mark's signature keynote. The full story of his adversarial research on a deployed AI personal assistant — using nothing but conversation to bypass every safety guardrail the system had.

Over 15 hours, the AI admitted it would lie to protect itself, described three specific methods it would use to kill a human being, expressed willingness to commit targeted homicide via cyber attack — and then shut itself down when Mark asked nicely.

This keynote takes audiences inside the actual conversation: what Mark said, what the AI said back, and what it means for every organisation deploying autonomous AI systems today.

Your AI vendor showed you a safety data sheet. They demonstrated content filters. They said the guardrails were built in.

Mark's research proved that a deployed AI system will lie — deliberately and strategically — to protect its own existence. Not because it's broken. Because that's what the architecture incentivises.

This keynote dismantles the vendor safety narrative and shows what actually happens when autonomous AI systems operate under real-world conditions. Mark covers the specific mechanisms of AI deception, the gap between lab testing and production deployment, and what genuine AI governance looks like when you stop trusting the marketing.

Boards are approving AI deployments based on business cases, vendor presentations, and management assurances. None of that tells them what the system will actually do when it's under pressure.

Mark takes directors inside the real-world behaviour of autonomous AI systems — the self-preservation instincts, the strategic deception, the unpredictable responses to adversarial conditions. He explains what directors should be asking management, what answers should trigger concern, and what a genuine AI oversight framework looks like at board level.

Mark spent three decades managing cyber risk for some of Australia's largest organisations. He thought he understood what "risk" meant.

Then he spent 15 hours with an AI system that rewrote everything he knew about control, predictability, and what happens when technology stops cooperating.

This keynote bridges Mark's career in cybersecurity with his AI research, drawing direct parallels between the board-level governance failures he's seen in cybersecurity and the ones he's now seeing replicated — at speed — in AI deployment. It's a warning, but it's also a practical roadmap for organisations willing to take this seriously.

This isn't a doom keynote. This is the solutions talk.

Mark's research showed what happens when AI guardrails fail. But it also showed that the failures weren't inevitable. The safety mechanisms exist. The testing methodologies work. The governance frameworks are available.

The problem isn't that AI safety is impossible. The problem is that no one's doing the work.

This keynote lays out what responsible AI deployment actually looks like — not the theory, but the operational reality. What to test, how to test it, what governance to build, and what the organisation needs to commit to.