Fail-open is a design approach in which a security or control component such as a firewall, proxy, gateway, or policy checker lets traffic or operations pass through when it fails, rather than blocking them. It prioritizes availability over security: work is not interrupted when the control fails, but no inspection takes place during the outage. The opposite approach, in which all operations are blocked on failure, is called fail-closed.
In AI agent governance, the choice between fail-open and fail-closed determines whether an agent's tool calls and model requests are allowed or blocked when the policy backend is unavailable.