Sydney AI Security & Governance Summit 2027

Beat the rush and join us early for complimentary barista-made coffee and breakfast.

Generative and agentic AI have moved from experimentation to production faster than most security and governance functions can keep pace with, and the organisations getting this right are treating security and trust as design requirements, not afterthoughts bolted on post-launch.

This opening keynote frames the day: what "trustworthy AI" actually requires in practice, and the shape of the threats and expectations now converging on organisations running AI at scale.

• Why AI trust and resilience are now board-level priorities, not just a technical concern.
• The forces reshaping AI risk right now: attackers, regulators, shadow adoption, and agentic autonomy.
• What secure-by-design actually looks like for production AI systems.

Regulators are sharpening supervisory expectations around AI faster than most governance frameworks can be rewritten, leaving organisations navigating real compliance obligations without a single, settled rulebook.

This keynote cuts through it: what's actually being asked of organisations right now, where enforcement is heading, and how to build governance that won't need rebuilding every time expectations shift.

• What regulators are actually asking of organisations deploying AI today.
• How voluntary and emerging standards are likely to harden into firm expectation.
• Designing an AI governance framework that flexes as the regulatory picture evolves.

Threat actors have moved from experimenting with AI to industrialising it, using it to scale phishing, write malware, and compress attack timelines from weeks to hours.

This panel brings together practitioners tracking this shift to unpack what's actually changed in attacker tradecraft, and what defenders need to do differently as a result.

• How AI has compressed attacker dwell time and reconnaissance-to-breach timelines.
• Real examples of AI-generated malware, phishing, and social engineering at scale.
• Where defensive AI can realistically keep pace with offensive AI, and where it can't.

Most AI security failures don't happen in the lab; they happen when a model moves into production and inherits real users, real data, and real attack surface.

This case study follows a team that built security controls around a live production AI system, covering model access controls, output validation, monitoring, and incident response, all retrofitted without stalling the business case for AI.

• What broke first when the model went from pilot to production scale.
• Building access control, logging, and monitoring around a live model without slowing deployment.
• The incident response playbook they wish they'd had on day one.

A hands-on, interactive session working through a real AI security and governance scenario as a room. Details announced soon.

Employees adopted generative AI tools faster than any policy could keep up, and most security teams are now flying blind on where sensitive data is actually going, both through unsanctioned tools and third-party AI dependencies buried in the supply chain.

This case study details how one organisation moved from having no visibility to a working discovery, risk-tiering, and sanctioned-alternative program.

• How they discovered the true scale of unsanctioned AI and third-party AI dependency use.
• Risk-tiering AI tools and vendors instead of a blanket ban that drives usage further underground.
• Building a sanctioned pathway that made compliance the path of least resistance.

Prompt injection remains the leading cause of agentic AI security failures in production, and as agents gain the ability to act, not just respond, the blast radius of a successful injection grows sharply.

This case study walks through a real prompt injection incident against an agentic system, what it exposed about the architecture, and the guardrails built afterward.

• Anatomy of a real prompt injection attack against a production AI agent.
• Why prompt injection is proving structurally difficult to fully patch, not just fix once.
• The guardrails, sandboxing, and human-in-the-loop checkpoints added after the incident.

Beyond prompt injection, AI systems face threats aimed at the model and the data itself, adversarial inputs designed to fool classifiers, poisoned training data, and deepfakes now sophisticated enough to bypass biometric and verification controls.

This panel examines the technical attack surface of AI systems and the defences that are actually holding up.

• How adversarial attacks and data poisoning differ from application-layer AI risks like prompt injection.
• What's actually detecting deepfakes and synthetic media in production today, and what isn't.
• Practical model-hardening steps that don't require a dedicated research team.

Small-group, discussion-based sessions where you'll work through real AI security and governance challenges with peers in similar roles. Roundtable topics will be announced soon.

Put your knowledge to the test in this fast-paced quiz covering real-world trivia, key concepts, and emerging trends. Compete for bragging rights — and a voucher — as the top scorer takes the crown.

AI agents now hold credentials, call APIs, and access systems on behalf of users, often with far broader permissions than anyone intended, and little visibility into what they're actually touching.

This case study covers how one organisation brought AI agent identities under proper access governance, closing shadow access gaps before they became breach headlines.

• Discovering the true scope of access AI agents had quietly accumulated.
• Applying least-privilege and identity governance principles to non-human, autonomous actors.
• Building ongoing visibility so new agents don't recreate the same shadow access problem.

Boards are being asked to oversee AI risk without always having the technical grounding to interrogate it properly, and regulators are starting to expect documented accountability, not just good intentions.

This closing keynote translates AI security and governance risk into the language and cadence boards actually need.

• What a genuinely useful AI risk report looks like at board level.
• Where accountability for AI risk should sit, and why "everyone" usually means no one.
• Building board AI literacy without turning every meeting into a technical briefing.

This interactive session puts the day's themes directly to the room with live voting, surfacing where the audience is genuinely exposed versus where the conversation has been focused, then debating the gaps in real time.

1. What's your organisation's single biggest AI security blind spot right now: shadow AI, agentic access, model-layer attacks, or third-party AI tools?
2. Who currently owns AI risk accountability in your organisation: security, risk and compliance, a dedicated AI governance function, or no one clearly?
3. Has your organisation experienced a prompt injection or agentic AI security incident, even a near-miss?
4. Is your board asking the right questions about AI risk: too little, about right, or overcorrecting into technical detail they can't act on?
5. Where will you spend your next AI security dollar: production model security, shadow AI discovery, agent identity governance, or model and deepfake defence?

Unwind with your peers for a couple of drinks on us!

Beat the rush and join us early for complimentary barista-made coffee and breakfast.

Generative and agentic AI have moved from experimentation to production faster than most security and governance functions can keep pace with, and the organisations getting this right are treating security and trust as design requirements, not afterthoughts bolted on post-launch.

This opening keynote frames the day: what "trustworthy AI" actually requires in practice, and the shape of the threats and expectations now converging on organisations running AI at scale.

• Why AI trust and resilience are now board-level priorities, not just a technical concern.
• The forces reshaping AI risk right now: attackers, regulators, shadow adoption, and agentic autonomy.
• What secure-by-design actually looks like for production AI systems.

Regulators are sharpening supervisory expectations around AI faster than most governance frameworks can be rewritten, leaving organisations navigating real compliance obligations without a single, settled rulebook.

This keynote cuts through it: what's actually being asked of organisations right now, where enforcement is heading, and how to build governance that won't need rebuilding every time expectations shift.

• What regulators are actually asking of organisations deploying AI today.
• How voluntary and emerging standards are likely to harden into firm expectation.
• Designing an AI governance framework that flexes as the regulatory picture evolves.

Threat actors have moved from experimenting with AI to industrialising it, using it to scale phishing, write malware, and compress attack timelines from weeks to hours.

This panel brings together practitioners tracking this shift to unpack what's actually changed in attacker tradecraft, and what defenders need to do differently as a result.

• How AI has compressed attacker dwell time and reconnaissance-to-breach timelines.
• Real examples of AI-generated malware, phishing, and social engineering at scale.
• Where defensive AI can realistically keep pace with offensive AI, and where it can't.

Most AI security failures don't happen in the lab; they happen when a model moves into production and inherits real users, real data, and real attack surface.

This case study follows a team that built security controls around a live production AI system, covering model access controls, output validation, monitoring, and incident response, all retrofitted without stalling the business case for AI.

• What broke first when the model went from pilot to production scale.
• Building access control, logging, and monitoring around a live model without slowing deployment.
• The incident response playbook they wish they'd had on day one.

A hands-on, interactive session working through a real AI security and governance scenario as a room. Details announced soon.

Employees adopted generative AI tools faster than any policy could keep up, and most security teams are now flying blind on where sensitive data is actually going, both through unsanctioned tools and third-party AI dependencies buried in the supply chain.

This case study details how one organisation moved from having no visibility to a working discovery, risk-tiering, and sanctioned-alternative program.

• How they discovered the true scale of unsanctioned AI and third-party AI dependency use.
• Risk-tiering AI tools and vendors instead of a blanket ban that drives usage further underground.
• Building a sanctioned pathway that made compliance the path of least resistance.

Prompt injection remains the leading cause of agentic AI security failures in production, and as agents gain the ability to act, not just respond, the blast radius of a successful injection grows sharply.

This case study walks through a real prompt injection incident against an agentic system, what it exposed about the architecture, and the guardrails built afterward.

• Anatomy of a real prompt injection attack against a production AI agent.
• Why prompt injection is proving structurally difficult to fully patch, not just fix once.
• The guardrails, sandboxing, and human-in-the-loop checkpoints added after the incident.

Beyond prompt injection, AI systems face threats aimed at the model and the data itself, adversarial inputs designed to fool classifiers, poisoned training data, and deepfakes now sophisticated enough to bypass biometric and verification controls.

This panel examines the technical attack surface of AI systems and the defences that are actually holding up.

• How adversarial attacks and data poisoning differ from application-layer AI risks like prompt injection.
• What's actually detecting deepfakes and synthetic media in production today, and what isn't.
• Practical model-hardening steps that don't require a dedicated research team.

Small-group, discussion-based sessions where you'll work through real AI security and governance challenges with peers in similar roles. Roundtable topics will be announced soon.

Put your knowledge to the test in this fast-paced quiz covering real-world trivia, key concepts, and emerging trends. Compete for bragging rights — and a voucher — as the top scorer takes the crown.

AI agents now hold credentials, call APIs, and access systems on behalf of users, often with far broader permissions than anyone intended, and little visibility into what they're actually touching.

This case study covers how one organisation brought AI agent identities under proper access governance, closing shadow access gaps before they became breach headlines.

• Discovering the true scope of access AI agents had quietly accumulated.
• Applying least-privilege and identity governance principles to non-human, autonomous actors.
• Building ongoing visibility so new agents don't recreate the same shadow access problem.

Boards are being asked to oversee AI risk without always having the technical grounding to interrogate it properly, and regulators are starting to expect documented accountability, not just good intentions.

This closing keynote translates AI security and governance risk into the language and cadence boards actually need.

• What a genuinely useful AI risk report looks like at board level.
• Where accountability for AI risk should sit, and why "everyone" usually means no one.
• Building board AI literacy without turning every meeting into a technical briefing.

This interactive session puts the day's themes directly to the room with live voting, surfacing where the audience is genuinely exposed versus where the conversation has been focused, then debating the gaps in real time.

1. What's your organisation's single biggest AI security blind spot right now: shadow AI, agentic access, model-layer attacks, or third-party AI tools?
2. Who currently owns AI risk accountability in your organisation: security, risk and compliance, a dedicated AI governance function, or no one clearly?
3. Has your organisation experienced a prompt injection or agentic AI security incident, even a near-miss?
4. Is your board asking the right questions about AI risk: too little, about right, or overcorrecting into technical detail they can't act on?
5. Where will you spend your next AI security dollar: production model security, shadow AI discovery, agent identity governance, or model and deepfake defence?

Unwind with your peers for a couple of drinks on us!