An attack path is a chain of access steps an attacker can take from an entry point to a valuable target, such as production systems or sensitive data.
Each step in an attack path can be legitimate on its own: a valid login, an approved OAuth grant, a role that was assigned on purpose. The risk comes from how the steps connect. Security tools that look at one permission at a time see a list of findings. Attack path analysis connects them and shows which ones end at something that matters.
For AI agents, the entry point is often the agent itself or a credential it holds. See the full guide to AI agent attack paths.
An attack path is a chain of access steps an attacker can take from an entry point to a valuable target, such as production systems or sensitive data.
The parts that make it up, in an environment with AI agents.
Where the path starts: an AI agent, a token, a service account or an OAuth app.
What that identity holds or can read: keys in files, tokens in CI, grants on its account.
Roles and scopes those credentials unlock, including roles that can be assumed.
Sensitive data or a destructive action in production at the end of the chain.
A real path shape, drawn the way the Elmoz security graph shows it.
Short answers about Attack path.
A vulnerability is a flaw in software. An attack path can be made entirely of valid access: a real token, an assigned role, an approved grant. Patching does not close it. Changing access does.
Remove or scope one step in the chain. The best choice is the breakpoint: the single change that closes the path with the least disruption to agents and people who depend on it.
Mostly they use access that already exists. What changes is speed and control: an agent can walk a path in seconds and can be steered by prompt injection, so a path that looked theoretical becomes usable.
Terms that belong to the same picture.
Book a demo and we walk through the access paths in an environment like yours.