Webinar Description
Key Takeaways
- Focuses on MITRE ATLAS, a framework mapping adversarial threats targeting AI systems across models, data, infrastructure and deployment pipelines
- Features updates from the Secure AI project, a collaborative research initiative led by MITRE’s Center for Threat-Informed Defense
- Includes perspectives from Fujitsu, Ensign InfoSecurity and Fortinet on applying ATLAS in operational security programmes
- Designed for security engineers, AI practitioners, threat intelligence teams and risk managers responsible for AI deployments
- Addresses the growing challenge of securing AI systems against targeted adversarial techniques
Introduction
As organisations accelerate their adoption of artificial intelligence, the attack surface for these systems has expanded considerably. Adversaries now target not only traditional IT infrastructure but also the models, training data, supply chains and deployment pipelines that underpin AI capabilities. This webinar, titled “Advancing MITRE ATLAS Through Collaborative R&D,” examines how the security community is responding to these emerging threats through structured frameworks and collective research efforts.
The session centres on MITRE ATLAS (Adversarial Threat Landscape for Artificial Intelligence Systems), a knowledge base that catalogues the tactics, techniques and procedures adversaries use to compromise AI systems. Hosted by AttackIQ in collaboration with MITRE’s Center for Threat-Informed Defense, the webinar provides security professionals with practical insights into defending AI-integrated environments using threat-informed approaches.
About This Event
This live virtual webinar brings together researchers and practitioners to discuss the current state of AI security and the collaborative work being undertaken to strengthen defensive capabilities. The programme combines expert-led presentations with technical discussion, offering attendees both conceptual grounding and operational relevance.
A central focus is the Secure AI project, an initiative launched by MITRE’s Center for Threat-Informed Defense that convenes industry members to contribute real-world observations and research to the ATLAS framework. The webinar presents findings from the latest Secure AI release and explores how participating organisations are integrating ATLAS into their security operations.
Understanding MITRE ATLAS and Its Role in AI Defence
MITRE ATLAS extends the threat-informed defence methodology that MITRE established with ATT&CK, applying similar principles to the distinct challenges of securing AI systems. Where ATT&CK documents adversary behaviour against enterprise networks and endpoints, ATLAS maps the specific techniques used to attack machine learning models, manipulate training data, exploit inference APIs and compromise AI supply chains.
The framework provides security teams with a common vocabulary for describing AI-specific threats and a structured approach to identifying defensive gaps. By cataloguing observed adversary behaviours, ATLAS enables organisations to move beyond generic security controls toward defences calibrated to the actual techniques being used in the wild.
This threat-informed approach proves particularly valuable as AI systems become embedded in critical business processes and infrastructure. Traditional security tools and methodologies were not designed with machine learning pipelines in mind, leaving organisations to address novel attack vectors without established playbooks. ATLAS helps bridge this gap by translating adversarial research into actionable defensive guidance.
The Secure AI Project and Collaborative Research
The Secure AI project represents a collaborative model for advancing AI security knowledge. Operated through MITRE’s Center for Threat-Informed Defense, the initiative brings together industry members who contribute observations, research and practical experience to continuously refine and expand ATLAS.
This collaborative approach addresses a fundamental challenge in AI security: the rapid evolution of both AI capabilities and the techniques used to attack them. Individual organisations often lack the breadth of visibility needed to understand the full threat landscape. By pooling knowledge across sectors and geographies, the Secure AI project accelerates the identification of emerging attack patterns and the development of corresponding defences.
The webinar highlights contributions from three participating organisations—Fujitsu, Ensign InfoSecurity and Fortinet—each bringing distinct perspectives shaped by their operational environments and client bases. These case studies illustrate how ATLAS translates from a reference framework into practical security programme enhancements.
The Expanding AI Threat Landscape
The urgency surrounding AI security reflects the growing sophistication and frequency of attacks targeting machine learning systems. Adversaries have demonstrated capabilities ranging from data poisoning—where malicious inputs corrupt model training—to model extraction attacks that steal proprietary algorithms through carefully crafted queries. Evasion techniques allow attackers to manipulate inputs in ways that cause models to produce incorrect outputs, potentially with serious consequences in applications such as fraud detection, content moderation or autonomous systems.
Supply chain vulnerabilities present additional concerns. Many organisations rely on pre-trained models, third-party datasets and open-source machine learning libraries, each introducing potential points of compromise. An attacker who successfully poisons a widely used training dataset or embeds a backdoor in a popular model can affect numerous downstream deployments.
These threats are not theoretical. Security researchers and threat intelligence teams have documented real-world incidents across multiple industries, prompting increased attention from both technical practitioners and executive leadership. The challenge lies in translating awareness into effective defensive measures—precisely the gap that frameworks like ATLAS aim to address.
Who Should Attend
The webinar is designed for professionals responsible for securing AI systems or integrating AI capabilities into existing security programmes. This includes security engineers working directly with machine learning infrastructure, AI engineers concerned with the integrity of their models and data, and cybersecurity researchers tracking adversarial developments in the AI domain.
Security operations and threat intelligence teams will find value in understanding how AI-specific threats fit within broader adversary campaigns. CISOs, security architects and IT risk managers can gain strategic perspective on the evolving risk landscape and the frameworks available to address it.
Organisations deploying AI in critical infrastructure, financial services, healthcare or other sensitive domains face heightened stakes and may find the threat-informed defence methodology particularly relevant. Cybersecurity vendors developing protective capabilities for AI systems can benefit from the research insights and collaborative opportunities presented.
Bridging Research and Operational Security
One of the persistent challenges in AI security has been the distance between academic research and operational practice. Researchers publish findings on novel attack techniques, but translating those findings into defensive measures that security teams can implement requires additional work. Frameworks like ATLAS serve as a bridge, encoding research insights in formats that align with how security programmes are structured and measured.
The webinar explores this translation process, examining how organisations are operationalising ATLAS within their existing security architectures. For teams already familiar with MITRE ATT&CK, the conceptual alignment between the two frameworks can accelerate adoption, allowing them to extend threat-informed practices to AI systems without building entirely new methodologies.
This practical orientation distinguishes the session from purely academic treatments of AI security. Attendees can expect discussion grounded in implementation realities, including the organisational and technical challenges of securing AI at enterprise scale.
Conclusion
As AI systems assume greater responsibility within organisations, securing them against adversarial threats becomes a business imperative rather than a research curiosity. This webinar offers security professionals an opportunity to deepen their understanding of the AI threat landscape and the collaborative efforts underway to address it. Through the lens of MITRE ATLAS and the Secure AI project, attendees can explore how threat-informed defence principles apply to the unique challenges of protecting machine learning systems in production environments.