A Case Study of Vulnerabilities in AI Agents

Virtual event

Feb 19, 2026 5:20 PM

Register today

Key Takeaways

Map real vulnerability classes in AI agent deploymentsWalk through actual prompt injection and broken authentication patterns that appear in production AI agents, with technical detail on how each one surfaces.
Recognize how attackers chain AI agent weaknessesSee how individual misconfigurations in authentication, input handling, and tool access combine into exploitable attack paths across agent architectures.
Build agent architectures that resist common exploit patternsLearn concrete design decisions for isolating agent permissions, validating inputs, and structuring guardrails that hold up under adversarial testing.

About this webinar:

AI agents are moving into production faster than most teams can secure them. This session breaks down real vulnerability classes found in AI agent deployments, from prompt injections to broken authentication, and shows what each one looks like in practice through detailed case studies from production applications and open-source repositories. Whether you are building agents internally or evaluating third-party tools, you will walk away with a clearer picture of where AI agent security stands today and what practical steps your team can take to stay ahead, including:

See real-world examples of prompt injection and broken authentication in AI agent systems
Understand how vulnerability classes interact to create exploitable attack paths
Take away specific architectural patterns for building more secure AI agents
Learn which agent design decisions reduce your exposure before deployment

For those who are:

Security engineers and AppSec practitioners who need to assess and harden AI agent deployments across their organization.
Platform and infrastructure teams who are building or integrating AI agents and want to understand the security tradeoffs before shipping to production.