Exploiting Copilot Studio's newest feature and exploring protection options
A deep dive into activation space of prompts in safety classifiers. Showing not why - but where - safety fails in LLM classifiers meant to detect malicious prompts.
An in-depth examination of emerging risks and effective mitigation techniques for protecting AI agents operating within the Bedrock AgentCore ecosystem.
A deep dive into realistic threat scenarios and practical strategies for securing enterprise AI agents built in Microsoft Foundry.
How adding a single safety labeled tool to an LLM's toolset can sharply increase its defense
0-click indirect prompt injection with tool use - a look through attribution graphs
How using structured prompts present findings of self-modeling in LLMs, which may benefit both attackers and defenders
How controlling the structure of the prompt, not just the semantics, can exploit your AI agents and their tools
