What happened
Arcjet (the security platform that ships inside AI code) launched agent runtime security for production AI agents, adding observability across agent workflows, deterministic policy enforcement before/after LLM, tool, database and API calls, and audit trails for security reviews and compliance. Announced Sept 17, 2026 (press release corrected Sept 18).
Why it matters
Adds a much-needed 'observe-enforce-audit' runtime control plane for agents that are already running in production, letting engineering teams govern consequential agent actions and giving security teams evidence for compliance — filling a gap between build-time guardrails and post-hoc logs.
Applicability
Engineering and security teams shipping agentic AI apps in production should evaluate Arcjet for runtime tool-call policy enforcement and audit evidence; assess now as agent deployments scale.