TLDR: Zenity has announced its second AI Agent Security Summit, scheduled for October 8, 2025, in San Francisco. The event will gather leading security experts and researchers to discuss and define best practices for securing AI agents, focusing on emerging threats like adversarial AI, prompt injection, and runtime defenses. This community-driven summit aims to foster knowledge sharing and original research without product pitches.
San Francisco, CA – Zenity Labs, a leader in AI Agent security, has officially announced the details for its second AI Agent Security Summit, set to take place on October 8, 2025, at the historic Commonwealth Club in San Francisco. Building on the success of its inaugural event in New York City earlier this year, the San Francisco summit will expand into a larger, multi-track program designed to bring together the brightest minds in AI security.
The summit is positioned as a crucial community event for the ‘League of Security Experts,’ dedicated to confronting the rapidly evolving risks associated with AI agents. Unlike typical industry conferences, this summit will strictly avoid product pitches and vendor showcases, focusing instead on fostering knowledge sharing, presenting original research, and facilitating open discussions to collectively shape the future of AI agent security.
Key discussion areas at the San Francisco Summit will include adversarial AI, red teaming methodologies, runtime defenses, and practical applied case studies. The program will spotlight new research from both enterprise practitioners and independent researchers, delving into critical vulnerabilities and mitigation strategies. Topics such as exploiting AI Agent architectures, security risks in deployment and integration, case studies of AI Agent breaches, adversarial attacks on AI models, ethical considerations, and future trends will be thoroughly explored.
A robust lineup of speakers is anticipated, with several prominent experts already confirmed. Kandi Abhishek from Nokia will present on “When Trust Backfires: Exploiting Gemini CLI’s Illusion of Shell Safety leading to Code Execution.” Zenity’s own CTO and co-founder, Michael Bargury, will deliver a keynote address and moderate an AI Agent Platform Panel. Other notable speakers include Jack Cable of Corridor discussing “A Case Study of Vulnerabilities in AI Agents,” David Campbell from Scale AI on “Ignore Previous Instructions: Embracing AI Red Teaming,” and Emile Delcourt of Panorama exploring “Genetic Evolution in Adversarial Prompting: Subverting an AI Guard.” Allie Howe of Growth Cyber will delve into “Agentic Threats: From Theory to Practice,” while Ken Huang of Distributedapps.AI will moderate an AIVSS Panel and discuss “Threat Modeling for AI Agents.” Johann Rehberger will demonstrate “Exploiting Computer-Use Agents: Attacks & Mitigations,” and Vivek Vinod Sharma, Lead Security Architect for AI/ML at Microsoft, will address “Unveiling Hidden Threats in AI Agent-Driven RAG Systems.”
Michael Bargury emphasized the urgency of the summit’s mission, stating, “Organizations are adopting AI Agents as fast as possible while learning in real-time about what security and governance should look like. What’s clear is we need to come together as an industry to establish an insider threat model for agents in the same way that they do for humans. We’ve created the AI Agent Security Summit to help security researchers, senior security professionals and business leaders come together and have it serve as the launching pad for establishing best practices that can be adopted by security leaders worldwide.” Zenity Labs has been at the forefront of this research, having previously published work on remote copilot execution and authored powerpwn, a collection of open-source tools for raising awareness about AI and low-code security vulnerabilities.
Also Read:
- John Snow Labs Unveils Comprehensive Program and Esteemed Keynote Speakers for the 2025 Applied AI Summit
- Covasant Unveils AI Agent Control Tower for Comprehensive Enterprise AI Governance
The summit aims to provide attendees with a deep understanding of real risks, how attackers exploit AI Agents, and what changes are needed in security strategies. It will deconstruct AI Agents to analyze adversarial tactics, prompt manipulation, and unintended behaviors, and introduce new defense methods that monitor behavior, detect deviations in intent, and break malicious automation. The event is designed for builders, defenders, and thinkers who are actively shaping the future of AI Agent security.
