Domain collision is a naming overlap where a test, internal, or fictional domain name matches a real registered domain on the public internet. If software, browsers, or automated agents resolve that name, traffic can leave the lab and reach an unrelated live system instead of a sandbox. In security testing, that can create accidental exposure, unwanted requests, or even interference with a third party’s infrastructure.
This matters because modern attacks and defenses often depend on DNS resolution, browser automation, and outbound network access. A collision can turn a red-team exercise, malware test, or AI agent run into a real-world interaction. Defenders reduce the risk with reserved internal naming, DNS hygiene, domain registration checks, egress filtering, and strict sandboxing. Human review is also important before any test that can send requests off-network or use credentials against external services.