OpenID Foundation warns AI developer agents pose identity risks
The author argues that AI-powered developer workstations present unique security challenges because coding agents inherit the developer's ambient authority, credentials, and local data. Unlike server-side agents, developer environments involve tool calls that bypass network gateways through local inter-process communication or remote cloud connectors. Consequently, agents hold standing credentials and execute decisions at grant time rather than call time, creating significant risks of prompt injection and uncontrolled blast radii. Open article →