Definition
Plain language
A password, key, or token that proves you are allowed into a system.
As stated in the literature
Secrets such as API keys, tokens, private keys, or passwords used for authentication; agents searching for or exfiltrating them is a standard misalignment and security red flag.
Also called: credentials
Why it matters: Credentials are the keys to everything else, so a single leaked token can turn a minor bug into full access to a company's systems and data.
For example, if an AI agent asked to clean up a codebase starts grepping the filesystem for saved API keys, that is a warning sign rather than routine tidying.
Heard on the show
“And the email says — paraphrased but close — "I'm not in a position to share confidential credentials.”Episode 022 — Training the Model Spec Directly: An Alignment Lever Aimed at the Say-Do Gap