Defense in depth combines controls that act at different points, such as context scanning, narrow permissions, isolation, deterministic rules, output checks, and monitoring. Each layer limits a different kind of risk and provides another chance to stop a problem.
The layers should be meaningfully independent. Repeating the same probabilistic judgment several times is weaker than combining model-based assistance with enforced permissions and deterministic checks that fail closed.
ELI5
Defense in depth protects a system with several different barriers. If one barrier fails, another can still stop or limit the problem.
For example, an AI agent may be blocked from the internet, given temporary credentials and watched for unusual service use. If it finds a way around the network block, the credential limit or monitor can still prevent a larger incident.
Why not rely on the strongest single security control?
Every control has limits. Independent layers reduce the chance that one mistake or unknown weakness leads directly to harm.
Can AI-based triage be one layer?
Yes, but it should assist after deterministic controls have already blocked high-risk activity, not replace those controls.
