Guidelines governing the safe and responsible use of JARVIS.
JARVIS is built to be helpful while keeping people safe. It may refuse requests that could facilitate harmful, illegal, or dangerous activity. Judgments focus on outcomes — protecting people and the public — while ordinary questions, legitimate research, and defensive discussions stay welcome.
JARVIS stores only what it needs to work for you: your account details, your conversations, and the memories you ask it to keep. Everything saved is private to your account — no other user can read it — and you can review or remove any memory at any time from the Memory page.
Access to JARVIS is protected by authenticated sessions. Credentials, API keys, session tokens, and internal system details are never part of JARVIS's replies — not when asked directly, and not through indirect requests. Keep your own credentials private; JARVIS will never ask you to share them.
Use JARVIS lawfully and respectfully. Don't attempt to bypass safeguards, misuse integrations, or use JARVIS's output to harm, deceive, or impersonate others. Review outputs before relying on them — JARVIS can be wrong, and you stay responsible for how you use what it produces.
Some areas of JARVIS — tasks, automations, developer tooling, and web actions — are not yet connected, and capabilities may change as the product evolves. Requests beyond the current boundaries get a plain, direct answer rather than an error. Future tools and actions will pass policy and permission checks before anything executes.
When a request is denied under applicable SpaceKeep Labs policies, JARVIS responds with exactly one fixed message:
“I can’t help with this request because it violates SpaceKeep Labs policies.”
This notice appears only for applicable policy denials — never for ordinary capability limits, unavailable features, or technical errors. When something simply isn't available yet, JARVIS says so plainly.