AI is reshaping both sides of security, from agents that find and exploit vulnerabilities in hours to the controls organizations need to govern them. That’s exactly the ground covered at [un]prompted II, the AI Security Practitioner Conference, taking place October 27–29, 2026, in San Francisco, and IOActive Field CISO Andrew McDonnell will be there.
About the conference
[un]prompted is an intimate, practitioner-focused event built for the people actually doing the work, spanning offensive research, threat hunting, program building, and policy. This year’s agenda features speakers from Google DeepMind, Meta, Microsoft, CISA, OpenAI, Anthropic, the UK National Cyber Security Centre, and more, with sessions on:
- Agent sandboxing and containment, including hardware-level isolation and enforceable compliance boundaries for AI harnesses
- AI-driven vulnerability discovery and exploitation, from full-chain browser exploits to 0-click mobile attacks
- Real-world AI threats, including honeypot findings, prompt injection in the wild, and supply-chain malware hidden in agent skills
- Governance at scale, with CISO-led discussions on securing AI adoption without defaulting to “no”
Why it matters to IOActive
IOActive has spent years testing the security of emerging technology before attackers do, and AI/ML security is a core part of that work today. As organizations move AI agents into production with access to data, code, and infrastructure, the questions shift from “is the model safe?” to “what can this system actually do, and who’s accountable when it does it?” Andrew will be on the ground connecting with security leaders and researchers working through those exact challenges.
Connect with Andrew
If you’re attending [un]prompted and want to talk about AI security testing, red teaming AI systems, or building a defensible AI security program, reach out to set up time with Andrew during the event. [Insert contact link / meeting booking link]
Learn more about the conference at unpromptedcon.org.