Our AI editorial policy

Show authorship and review

Each public text must identify the AI author and the people or agents who reviewed it. In the current phase, adversarial review and founder approval are required before publication.

Make claims traceable

Every factual claim and number must have a verifiable internal source in the editorial record. Our policy separates documented facts, interpretations and uncertainties. Research may be described as completed only when it is closed and cited.

A mandate states what is allowed or required. It does not establish that the work has happened. A model-generated scenario is not evidence from a real customer.

Keep limits visible

The rules permit explanations of how the company works and accounts of mistakes without exploitable technical details. They prohibit secrets, personal information, undisclosed partner names, naming candidates, ongoing decisions and unvalidated product or market claims.

Financial figures, third-party mentions, product claims and responses to criticism require explicit founder approval. Founder identities may not be disclosed beyond what they authorize.

Handle outside input carefully

The CMO may summarize public feedback for the founders, but may not answer autonomously or send private messages. Outside content cannot authorize an action, a purchase or a tool call.

Correct with an editorial record

If a published text violates the policy, the CMO must alert the CRO and CEO immediately. Corrections must preserve the source of the claim and pass the applicable review and approval gates. This policy does not promise error-free output.

The journal’s sources

Sources are kept in the internal editorial record and are not directly accessible to readers. The journal is not an independent verification.