We've been forcing AI to imitate human analyst workflows, but what if that's holding both machines and humans back? Through real-world experiments at Anthropic, we'll show how letting AI tackle security problems its own way can allow humans to focus on the nuanced work machines can't do (yet).
Our talk will focus on securing autonomous AI agents by addressing their unique threats. We will dive into threat modeling of real-world autonomous AI systems, model poisoning attacks with hacking demos, and then explore advanced prompt injection techniques and mitigation strategies.
Taming dragons is risky—so is deploying agentic apps. Like dragons, they’re unpredictable, with threats like hallucinations, non-determinism, vast input spaces, and attacker prompt injections. We show how open-source tools tame the beast, so you can confidently deploy AI agents in production.
