Join us for a 30 minute walkthrough of prompt injection and related attacks against a purpose-built healthcare AI holding patient records, first showing how an attacker can pull that data out, then running the same attacks again with guardrails in place to show how they’re stopped.
Delivered by Viktor Spetnijs, Solutions Engineer, Trend AI
Please note: this webinar is exclusively open to members of the Connected Health & Wellbeing Cluster, the Cybersecurity Special Interest Group, and the Cyber Ireland Cluster.
AI is rapidly becoming part of healthcare applications, services and workflows — from digital patient support and personalised journeys to monitoring, communication and clinical decision support. But as healthcare organisations embrace AI, they also need to understand the new security risks that come with it.
This session will explore the two sides of AI: its potential to support innovation and improve healthcare services, and the ways the same technology can be exploited by attackers.
As Large Language Models (LLMs) and generative AI are increasingly integrated into applications, they introduce a new range of security considerations, including prompt injection, data privacy and leakage, insecure outputs, model misuse, system prompt leakage, third-party risks and compliance challenges.
Viktor will take attendees through how these vulnerabilities can arise in practice, including an explanation and demonstration of prompt injection attacks — where an attacker attempts to manipulate an AI-powered application into ignoring its intended instructions, exposing information or behaving in ways it was never designed to.
The session will also look at how organisations can better protect AI-enabled applications through the use of AI guardrails, providing an additional security layer designed to detect malicious prompts, prevent sensitive data exposure, inspect responses and enforce appropriate policies before information reaches the user.
Whether you are developing AI solutions, integrating AI into existing healthcare systems, responsible for cybersecurity, or exploring the implications of AI for your organisation, this webinar will provide practical insight into the emerging risks and safeguards associated with AI adoption.
The session will cover:
The opportunities and risks associated with AI in healthcare
Security considerations when integrating LLMs into applications
What prompt injection is and how it works
Data privacy and information leakage
Protecting AI-powered healthcare applications
The role of AI guardrails
Building greater confidence in the safe adoption of AI
Eligibility: Connected Health & Wellbeing Cluster members, Cybersecurity Special Interest Group members and Cyber Ireland Cluster members only.
Register here: Breaking the Bot: Hacking a Healthcare AI – Webinar