Webinar Description
Key Takeaways
- Live webinar examining the Hugging Face security breach and its implications for AI system protection
- Focus on agentic threat modeling as a methodology for identifying risks across the AI lifecycle
- Practical guidance on governance, control definition, and secure-by-design principles for AI deployments
- Designed for security professionals, AI engineers, architects, and technology leaders
- Hosted by ThreatModeler with participation from Aedify Security
Introduction
Building Secure AI with Threat Modeling is a live webinar that examines how organisations can strengthen their approach to AI security following high-profile incidents in the machine learning ecosystem. The session is designed for security professionals, AI engineers, and technology leaders responsible for deploying and managing AI systems. With agentic architectures becoming increasingly prevalent across enterprise environments, the webinar addresses the growing need for structured risk identification methodologies that account for the unique characteristics of modern AI deployments.
Learning from the Hugging Face Security Incident
The webinar uses the recent Hugging Face security breach as a central case study, breaking down what occurred and extracting lessons applicable to organisations operating their own AI infrastructure. Hugging Face, as one of the largest repositories for machine learning models and datasets, represents a critical node in the AI supply chain. Security incidents affecting such platforms have cascading implications for downstream users who depend on shared models and components.
Rather than treating the breach as an isolated event, the session positions it as indicative of broader vulnerabilities inherent in how AI systems are currently developed and deployed. This framing helps attendees understand why traditional application security approaches may prove insufficient when applied to machine learning workloads.
Agentic Threat Modeling for AI Systems
A core focus of the webinar is the application of agentic threat modeling to AI security challenges. This methodology extends conventional threat modeling practices to address the specific risk profiles of agentic architectures, where AI systems operate with greater autonomy and interact with external resources, APIs, and data sources.
The session demonstrates how threat modeling can be applied throughout the AI lifecycle, from initial design through deployment and ongoing operations. This lifecycle perspective acknowledges that AI systems present evolving attack surfaces as models are updated, fine-tuned, or integrated with new data sources. Identifying risks at each stage enables organisations to implement controls proactively rather than reactively addressing vulnerabilities after deployment.
Governance and Control in AI Deployments
Beyond technical threat identification, the webinar addresses governance frameworks necessary for maintaining oversight of AI systems at scale. As organisations deploy multiple models across different business functions, establishing clear control boundaries becomes essential for both security and compliance purposes.
The discussion covers how to define control mechanisms that provide meaningful security assurance without impeding development velocity. This balance is particularly relevant for organisations operating in regulated industries such as finance and healthcare, where AI governance requirements continue to evolve alongside broader regulatory frameworks.
Who Should Attend
The webinar is structured for practitioners and decision-makers involved in AI security strategy and implementation. Security architects and CISOs will find value in the governance and risk management perspectives, while AI and ML engineers can apply the threat modeling techniques directly to their development workflows. Solutions architects working on AI integration projects and product leaders overseeing AI-powered features will benefit from understanding how security considerations should inform architectural decisions.
Organisations in technology, financial services, healthcare, manufacturing, and critical infrastructure sectors face particular urgency in addressing AI security given their regulatory environments and the sensitivity of data processed by their AI systems.
Secure-by-Design Principles for AI
The webinar reinforces secure-by-design principles as foundational to effective AI security programmes. This approach embeds security considerations into the earliest stages of AI system development rather than treating security as a final validation step. For agentic architectures specifically, secure-by-design practices help ensure that autonomous AI behaviours remain within intended boundaries and that appropriate safeguards exist for interactions with external systems.
ThreatModeler hosts the session with participation from Aedify Security, bringing together perspectives on how threat modeling tools and methodologies can support organisations in operationalising these principles across their AI portfolios.