Never paste secrets, customer data, or live IOCs

Before AI can help you, you have to decide what you are allowed to feed it. In a consumer chatbot, whatever you paste may leave your control and, depending on the terms, may be retained or used to improve a model. For a defender, that is a data-handling problem with teeth.

Never paste into a tool you do not control:

Secrets. API keys, passwords, tokens, connection strings, private certificates. Redact them or swap in placeholders.

Customer and personal data. Names, account numbers, health or financial records, anything covered by contract or regulation.

Live indicators tied to an active case. Real internal hostnames, user identities, or the specific indicators of a breach in progress. Leaking these can tip off an attacker or expose an investigation.

Safer habits that keep the help without the risk:

• Prefer an enterprise or private deployment that your organization has vetted, with data-retention and training terms in writing, over a personal account.
• Sanitize before you paste: replace real identifiers with tokens like HOST1 or USER1 while keeping the structure intact.
• Ask whether the tool runs in your own tenant or environment, and whether your data is used to train any model. Regulatory concern about training data is among the most cited barriers to AI in the SOC.

The rule of thumb: if a leak of this exact text would be an incident on its own, it does not go into a tool you do not fully control.