What happened
Ory announced general availability (Sept 10) of Ory Agent Security, described as the first Agent IAM control plane that evaluates an agent's tool-action against policy before execution, binding actions to a human owner and delegated agent identity. It supports 11 agent harnesses and 14 SDK frameworks, can run in observation mode before enforcement, and exports decisions via OpenTelemetry; deployable via Ory Network or self-managed/air-gapped.
Why it matters
Moves agent security from passive monitoring to pre-execution, revocable, least-privilege authorization at the tool boundary — directly addressing the harness/identity control-plane gap regulators (ASD, proposed Stop Rogue AI Act) are now calling out as the primary agentic-AI risk surface.
Applicability
Enterprises deploying production AI agents/harnesses should evaluate for agent inventory, owner-binding, and per-action policy enforcement; especially relevant for regulated or air-gapped environments.