Server-side request forgery, often shortened to SSRF, is a flaw that lets an attacker make a server send web requests on their behalf. The danger is that the server can usually reach places the attacker cannot, such as internal systems behind a firewall. Think of a receptionist who will fetch any file you name. You cannot walk into the back office, but you can ask the receptionist to bring something out of it.
The usual way in is a feature that fetches a web address for the user, like a link preview or a webhook. If the server fetches whatever address it is given without checks, an attacker can point it at internal services or at cloud settings that hold login credentials.
Why it matters
Many AI tools fetch web pages for you, read links you paste in, or connect to other services. Each of those features is a place where this flaw can appear. When you choose a tool, ask the vendor how its servers limit which addresses they will fetch. We list known problems in our advisories. Related problems include path traversal and remote code execution.
Published Updated