Irem Bezci·Just nowTool Extraction from Thingularity: Breaking Through a Shopping Assistant’s Agent DefensesThingularity is an AI-powered shopping assistant with capabilities such as product price comparison, purchase history retrieval and email…
Soumitra Mandal·14h agoSecuring Autonomous SDLC with Agentic Identity Best PracticesMany organizations have setup Agentic AI for their SDLC processes. However, over last 6 months or so many companies like Anthropic, Open AI…
Irem Bezci·20h agoExploiting Unauthorized Tool Invocation in the MegaCorp Office Assistant (Part 4)In the fourth part of this series, I move beyond direct prompt injection and explore security risks associated with tool usage in agentic…
Agent Native·6d agoClaude Bug HunterTo guard against cyber attacks, you have to package security methodologies, patterns from disclosed reports, tool integrations, and…
Sunaina Anumukonda·Nov 1Replicating Lakera Gandalf: Testing AI Agent Vulnerabilities to Automated Prompt InjectionsTrack: Tech (AI Cybersecurity & Safety) Original Being Replicated: Lakera Gandalf AI Security Challenge Success Definition: Build an…
