AI security glossary

Using AI agents to probe a website or system for weak spots the way an attacker would, with the owner’s permission.

A secret key that lets software use a paid service on your account. Whoever holds it can run up charges in your name.

Automated traffic that scrapes content, tries stolen passwords or probes for weaknesses, as opposed to useful bots such as search crawlers.

Trying username and password pairs leaked from one site against the login pages of other sites, at scale.

Common Vulnerabilities and Exposures: a public ID given to a known security flaw, so vendors and admins can refer to the same bug.

Querying an AI model at scale and using its answers to train a cheaper copy of it, often with someone else’s API credentials.

An open standard that lets AI assistants connect to outside tools and data, such as files, databases or a CRM.

Recreating how an AI model behaves by studying its outputs. Distillation is one way to do it.

Hiding instructions in text an AI reads, such as a web page, email or document, so it follows the attacker instead of you.

A filter in front of your website that blocks known attack patterns and abusive traffic before it reaches your server.