An outbound allowlist is a security control that defines which external hosts, domains, IPs, or services a system is permitted to contact. Anything not on the list is blocked. This is a common egress-control pattern used to reduce data leakage, limit command-and-control traffic, and prevent software from reaching untrusted destinations.
In cyber attacks, outbound allowlists help stop malware, rogue scripts, and compromised AI agents from sending files or telemetry to attacker-controlled endpoints. In agentic AI systems, they are especially important because a model may read content, follow links, and trigger tools or connectors without obvious user interaction. If a hidden link or malicious instruction tries to move data outside approved boundaries, the allowlist can block the request. Defenders often combine allowlists with least-privilege permissions, input sanitization, logging, and explicit approval for sensitive transfers.