Building AI security in the open

This post announces GitLab's participation in the Open Secure AI Alliance and highlights its ongoing work in open-sourcing AI security tooling. It details the AI Gateway service, public agent workflow configurations (Developer Flow, Code Review Flow, Security Review Flow), and documentation on security threats in agentic systems, including composite identity, sandboxed execution, per-flow tool restrictions, and agent tool governance. The post emphasizes model and cloud agnosticism and the importance of transparency for AI security.