Docker, Snyk, and Keycard Unveil Agent Baseline for Enterprise AI Security

Docker, Snyk, and Keycard have jointly developed Agent Baseline, a v1.0 draft blueprint defining 35 security controls across six outcomes for enterprise AI agent adoption. The framework addresses risks such as unchecked agent authority, malicious instructions, and runtime reprogramming by ensuring agents operate within approved boundaries. Key controls include discovering agent dependencies, constraining access, authorizing actions, observing operations, validating outputs, and responding to incidents. The blueprint was unveiled at Black Hat 2026 during the session 'Securing your AI Agent: The Road to Software Factory.' This initiative aims to provide enterprises with a standardized approach to managing AI agents securely as they transition from experimental to operational use.