Conference Description
Key Takeaways
- A six-hour virtual summit focused on securing AI-assisted software development, particularly for teams using Claude Code and similar coding agents
- Covers the full lifecycle from machine-readable security requirements through adversarial testing, trust boundary mapping and enterprise governance
- Faculty includes OWASP standards leaders, ethical hackers and enterprise security executives
- Live demonstrations showing specification creation, vulnerability exploitation, remediation verification and approval workflows
- Designed for developers, application security engineers, DevSecOps teams and engineering managers responsible for AI-assisted delivery
Introduction
The Claude Code Security Summit is a virtual event designed for developers and security practitioners navigating the operational challenges of AI-assisted software development. As coding agents evolve beyond simple code suggestions to autonomous systems capable of generating features, reviewing pull requests and remediating vulnerabilities, organisations face a fundamental question: how do teams establish and enforce security boundaries when AI can create, evaluate and act on software at machine speed?
This six-hour programme addresses that challenge through four connected sessions covering security requirements, adversarial validation, coding-agent controls and enterprise governance. The event is structured as vendor-neutral and practitioner-led, focusing on workflows and decisions rather than product demonstrations.
The Security Challenge with Autonomous Coding Agents
Most development teams have experimented with AI coding tools in some capacity. The more significant transition occurs when those tools gain the ability to inspect repositories, generate changes across multiple files, invoke external tools and produce work that appears ready for deployment. At this stage, prompting skill alone becomes insufficient.
Without enforceable requirements, security gates and clear human ownership, coding agents can reproduce insecure patterns, expose sensitive tools and data, and push unverified changes toward production at speeds that outpace traditional review processes. Teams require machine-readable requirements, scoped tasks, automated tests, security checks, adversarial validation, least-privilege tool access, audit evidence and human approval at decisions that carry genuine risk.
Programme Structure and Live Demonstrations
The summit follows a connected narrative: specify securely, build, attack and validate, control the coding agent, then govern enterprise adoption. Each session builds upon the previous, creating a comprehensive view of the security lifecycle for AI-assisted development.
Live demonstrations form a central component of the programme. Attendees will observe the creation and connection of specification files including REQUIREMENTS.md, ARCHITECTURE.md, SECURITY.md and CLAUDE.md for AI-assisted projects. Sessions will compare the same feature built with and without specifications to illustrate the security differential, then run model-led validation loops against explicit requirements.
The adversarial component includes attacking AI-generated APIs and applications, tracing findings to root causes and converting fixes into regression requirements. Coding-agent workflows using Model Context Protocol will be mapped across identities, tools, data, permissions and third-party trust boundaries, with abuse paths demonstrated alongside practical guardrails. Enterprise approval flows covering vulnerability remediation, threat modelling, documentation and human sign-off will also be examined.
Faculty
Jim Manico, Founder of Manicode Security, brings more than fifteen years of experience training developers, architects and security engineers. He leads several OWASP projects including AISVS, ASVS, the Cheat Sheet Series and Proactive Controls, and is a former OWASP Global Board member and Java Champion.
Vandana Verma, Security Relations Leader at Snyk and former OWASP Global Board member, contributes nineteen years of cybersecurity experience. She runs OWASP AppSec Days India and supports community initiatives including InfosecGirls, InfosecDiversity and InfosecKids.
Katie Paxton-Fear, Staff Security Advocate at Semgrep, is a former developer who specialises in breaking applications. She combines API and application security expertise with developer education, and her session on hacking systems from military to AI was recognised as a top session at RSA.
Michael Shost brings more than twenty-five years of enterprise transformation experience and created the Agentic AI-Driven Enterprise Delivery Model. He holds certifications as a Chief Information Security Officer and Certified Ethical Hacker.
Target Audience
The summit is built primarily for practitioners creating, testing and approving AI-assisted software. This includes software developers and application engineers using Claude Code or other coding agents, application security and product security engineers reviewing AI-generated code, DevSecOps and platform security teams implementing delivery controls, security engineers and ethical hackers validating AI-generated applications, and engineering managers responsible for AI-assisted delivery.
The programme also holds value for AI security engineers, coding-agent platform owners, CISOs, technology risk leaders, security consultants and governance professionals. Attendees should understand modern software delivery and be prepared to examine real workflows, code and security decisions, though expertise in every standard or agent framework is not required.
Practical Outputs
Attendees will leave with adaptable materials including requirements files, test gates, trust-boundary maps, threat models and approval checklists. The programme emphasises outputs that can be applied immediately to existing development workflows, enabling teams to define automated test, security and human-review gates required before generated code can progress, and to draft approval checklists stating acceptable autonomy, mandatory human decisions and minimum production evidence.