SpecterOps & OpenAI launch AI security training course

SpecterOps has launched an AI security training course in partnership with OpenAI as part of OpenAI's Daybreak Defence Network.

The course, Adversary Intelligence: LLM Tradecraft, is part of SpecterOps' Tradecraft Academy and is aimed at security practitioners working with large language models and agentic systems. It includes hands-on labs covering threat modelling, security testing, reverse engineering, and defensive security workflows.

Participants will also receive access to Codex and OpenAI's latest models. The material is designed to give practitioners practical experience in building, evaluating, testing, and defending AI systems used in organisations.

The launch comes as more businesses deploy AI tools in routine operations and security teams face new risks linked to machine-led interactions, automated tools, and additional system identities. The course is intended to address gaps in practitioners' understanding of how these systems work, where they can fail, and how attackers may approach them.

According to SpecterOps, the training includes eight hours of content and covers both offensive and defensive aspects of AI security. Topics include the architecture and limitations of LLM and agentic systems, common attack methods such as prompt injection and jailbreaks, and the use of AI in security testing and reverse engineering.

Wunan Li, Global Cyber Partnerships at OpenAI, described the thinking behind the programme.

"Building practical experience is essential to understanding how AI can be applied effectively in cybersecurity," Li said. "Built by SpecterOps' AI threat researchers using the latest OpenAI models, Adversary Intelligence: LLM Tradecraft gives security practitioners the skills and understanding they need to proactively manage changing enterprise environments."

SpecterOps has positioned the course within a wider training portfolio for cybersecurity professionals. Through Tradecraft Academy, it offers live and on-demand instruction, including courses on attack path management using BloodHound, software supply chain security, and Kubernetes for red teamers.

Course focus

The training centres on three main areas. The first is how to build and assess LLM and agentic systems, including reviewing their design, limitations, and security implications.

The second covers adversarial techniques aimed at AI systems, including prompt injection, jailbreaks, and weaknesses in AI infrastructure that attackers could exploit.

The third examines the use of AI in defensive security work, including threat modelling, security testing, and AI-assisted reverse engineering using OpenAI models.

Andrew Chiles, Vice President of Tradecraft at SpecterOps, said the aim was to help practitioners move beyond simply using AI tools toward a clearer understanding of their risks and weaknesses.

"Increasingly, security practitioners are being asked to work with LLMs and agentic systems, but using these technologies effectively requires understanding how they work, where they can fail, and how an adversary might approach them," Chiles said. "The gap between using AI and understanding it can create security blind spots. We built the Adversary Intelligence: LLM Tradecraft course to close that gap through hands-on experience, giving practitioners the foundation to evaluate these systems critically and make informed decisions to build, evaluate, and secure them."

SpecterOps is best known as the creator of BloodHound, a tool used to analyse identity-related attack paths in enterprise environments. Its broader work focuses on helping organisations understand identity risk across AI and hybrid systems.

The addition of an AI-focused course reflects a growing market for specialist security training as companies integrate generative AI and autonomous software agents into internal processes. For security teams, that shift has increased demand for practical instruction on how these systems behave, how they can be tested, and where they are exposed to misuse.

Participants will receive 30 days of access to the course content and a separate 30-day period of access to Codex.