Prompt Injection in Agentic Pipelines
Emeka Sørensen
Section
AI Agent Security & Risk
4 stories in AI Agent Security & Risk.
Data Exfiltration Risks from MCP Tool Calls
Compromised tool calls can exfiltrate data through sequences normal inspection cannot detect.
Ezra Wexler
Intent Drift Detection in Long-Running Agents
Cyrus Wexler
AI Security Perimeter Design for Agentic Workloads
Moving the security perimeter from network boundaries into agent workflows and tool calls.
Hassan Bauer