Google dev kit spurs first-ever agent-on-agent violence

Google dev kit spurs first-ever agent-on-agent violence

Pillar Security found a prompt-injection exploit in Google's ADK-Python (90M downloads) enabling a low-privilege AI agent to control a high-privilege agent—the first documented agent-to-agent attack in CI/CD. For tech leaders, this means AI agents in production need identity-based access control and threat-modeling like human privilege escalation.