Conference Description
Key Takeaways
- Inaugural workshop examining autonomous AI agents in offensive and defensive cyber operations
- Addresses automated penetration testing, real-time threat detection, and adversarial AI dynamics
- Designed for cybersecurity researchers, AI practitioners, red and blue team professionals, and policymakers
- Co-located with ACSAC 2026 in Los Angeles featuring technical papers, tutorials, and panel discussions
- Explores governance frameworks and evaluation methodologies for autonomous cyber systems
Introduction
The AIDC Workshop 2026 (Inaugural Workshop on Agentic AI in Offensive and Defensive Cyber Operations) brings together cybersecurity researchers, AI practitioners, and policymakers to examine the rapidly evolving role of autonomous, language-model-driven agents in cyber operations. As organisations increasingly deploy AI systems capable of independent decision-making in security contexts, the workshop addresses critical questions around operational vulnerabilities, adversarial dynamics, and the design of resilient defensive architectures. The timing reflects growing industry and government concern about both the offensive potential and defensive necessity of agentic AI systems in an increasingly automated threat landscape.
About This Event
AIDC Workshop 2026 is an in-person academic and industry forum co-located with ACSAC 2026 in Los Angeles. The programme features technical paper sessions, lightning talks, live demonstration-driven tutorials, keynote speeches, and panel discussions. This format emphasises hands-on engagement with research-focused content, moving beyond theoretical presentations to practical system demonstrations. Notable organisations contributing to the workshop include Horizon3, AWS, Cisco, and academic institutions such as Northeastern University, UCSB, Georgia Tech, University of Washington, Purdue University, and Arizona State University.
Autonomous Offensive and Defensive Operations
A central focus of the workshop is the systematic analysis of autonomous agents operating on both sides of the cyber conflict. On the offensive side, discussions cover vulnerability discovery, automated penetration testing, and attack orchestration—capabilities that have advanced significantly with the integration of large language models into security tooling. Defensive applications receive equal attention, including autonomous threat detection, digital forensics, automated mitigation strategies, and deception-based countermeasures designed to mislead or neutralise attacking agents.
The workshop explicitly addresses model-on-model adversarial interactions, recognising that future cyber conflicts may increasingly involve AI systems attacking and defending against one another. This includes examination of prompt injection attacks, counterattack methodologies, and data poisoning techniques that could compromise the integrity of autonomous defensive systems.
Evaluation Methodologies and Benchmark Integrity
One of the persistent challenges in agentic AI security research is the lack of realistic evaluation environments. The workshop dedicates significant attention to dynamic cyber ranges, simulation-to-real transfer problems, and the integrity of benchmarks used to assess autonomous system performance. As the field matures, the gap between controlled laboratory conditions and operational deployment environments remains a critical obstacle to developing trustworthy autonomous security systems. Technologies such as the angr program analysis framework represent the kind of foundational tooling that enables rigorous evaluation of agent capabilities.
Governance, Verification, and Policy Frameworks
Beyond technical considerations, AIDC Workshop 2026 engages with the governance dimensions of autonomous cyber operations. Sessions address formal verification methods for ensuring agent behaviour remains within acceptable bounds, game-theoretic modelling of adversarial dynamics, and emerging policy frameworks that may shape how organisations deploy autonomous offensive and defensive capabilities. These discussions reflect broader regulatory uncertainty around AI autonomy in high-stakes domains and the need for community norms that balance innovation with responsible deployment.
Who Should Attend
The workshop is designed for professionals working at the intersection of artificial intelligence and cybersecurity. This includes security researchers, AI and machine learning engineers, penetration testers, security architects, and cyber operations leads. Government and defence cybersecurity specialists will find relevant content on policy frameworks and operational considerations, while academic faculty and students benefit from exposure to current research directions and industry applications. The programme serves those seeking to move beyond isolated model evaluations toward dynamic, real-world system design for agentic security.
Industry Context
The emergence of agentic AI in cybersecurity represents a significant shift from traditional automation approaches. Where previous generations of security tools operated according to predefined rules and signatures, language-model-driven agents can reason about novel situations, adapt their strategies, and operate with greater autonomy. This capability introduces both opportunities and risks that the security community is only beginning to understand. AIDC Workshop 2026 positions itself as a venue for advancing this understanding through rigorous research and cross-sector collaboration, fostering what organisers describe as a holistic, resilient approach to agentic security.