Defending Agents From Prompt Injection Without a Bigger Model

​Prompt injection can turn a helpful agent into an unpredictable one, but the answer doesn’t have to be a bigger model. Join Opensciencelabs for a lively online session exploring practical ways to defend agents against malicious instructions and keep them on track.

​Bring your questions, curiosity, and a drink of choice. We’ll dig into approaches that make agent systems more resilient without simply throwing more compute at the problem.