Acceptable use

Version 2026-09-29b, effective 2026-09-29

This document has not yet been reviewed by a lawyer. It describes how ZenOS actually works today; the provider's full registered details will be added before paid plans open.

What this workspace is for

This product is for running agents against your own company's work — your projects, your clients, your billing, your outreach. It is not a general-purpose tool for actions unrelated to running your agency.

What you may not use it for

• Sending unsolicited messages outside the limits and consent your plan allows.
• Impersonating a person or another business, including through an agent's output.
• Feeding an agent someone else's confidential material without the right to do so.
• Attempting to make an agent bypass the approval it is required to get before an action that changes the outside world.
• Using the product to build a competing product from the inside.
• Uploading special-category data (health, ethnicity, religion, biometrics, sexual orientation) or data about criminal offences. The product has no specific safeguards for these categories, and doing so is a breach of this policy regardless of your reason.

Contact-related work — the stricter rules

Two kinds of action carry stricter rules than the rest of the product: looking up details about someone who has never contacted you (contact enrichment), and sending that person a first-contact message (cold outreach). Both are kinds of action this product recognises by name and treats specially.

Contact enrichment and cold outreach are not available today. If either is ever offered, everything below applies to it from its first day; the pricing page shows whether each is available.

• You must have your own lawful basis before you ask an agent to do either — typically legitimate interest in a business-to-business context, assessed against the recipient's own jurisdiction's rules on unsolicited business contact (for example, the ePrivacy rules in the EU and the UK, and the CAN-SPAM Act in the US). We do not establish or verify that basis for you; you are the one making the contact, and you remain responsible for it.
• Region limits. We do not knowingly enable cold outreach into a jurisdiction where it would require a form of consent or registration we have not implemented (for example, jurisdictions requiring prior opt-in for business email). If you are unsure whether a region you intend to contact is covered, ask us before you rely on the feature.
• A daily limit. Each agent can carry a daily limit on how many people it may contact, and the approval checks refuse anything beyond it — so a misconfigured or compromised workspace is stopped before it can do much damage.
• A person approves every lookup and every message first, permanently. Every contact lookup and every first-contact message waits in your approval queue until a person in your workspace approves it, before anything happens. No plan, no autonomy setting and no amount of an agent's approval history changes that, for either kind of action. It is the same permanent rule that applies to issuing an invoice and publishing to a public account.
• A do-not-contact list. Each workspace has its own do-not-contact list. Anyone in your workspace can add an email address or a company's whole domain to it, and it applies at once, without waiting for an administrator. The product checks the list before a lookup, when a person reviews a contact, when a message is approved and again just before it is sent, and refuses anything on it. Someone who uses the stop link in a message is added to the list automatically.
• A notice the product adds, not the agent. After the agent's draft, the product itself adds a notice to every first-contact message: first, on its own, the recipient's right to object; then who is contacting them and how to reach you, why, on what basis, which of their details are held and where they came from, how long they are kept, their rights (including the right to complain to the data protection authority), and a link to your privacy notice. Every message also carries a working link to stop further messages. The agent cannot write, change or leave out any of this. If the notice cannot be completed — for example, because your workspace has not recorded where its privacy notice is published — nothing is sent.

Our position on the EU AI Act

ZenOS is designed so that its AI agents assist a business's own internal operations (reporting, drafting, and similar support tasks), and so that the system itself never takes an action with legal or similarly significant effect without a human decision in the loop — every action that changes something outside your workspace is a proposal a person or a declared policy must approve. On that design basis, we intend the core product not to be a prohibited AI practice under the EU AI Act, and we do not currently offer any of the specifically high-risk use cases the Act lists (for example, biometric identification, employment decision-making, or credit scoring). If you intend to use agent output as an input to a decision the Act would treat as high-risk, that use is outside what this product is designed or represented to support, and you are responsible for your own compliance assessment before doing so. This section describes our design intent; a lawyer's review of this section is pending, and we will update this position as our own use of the product and the Act's implementing guidance develop.

Where the product shows you text a model wrote — an agent run's report, or a proposal waiting in your approval queue — it is labelled "Written by an AI agent" right next to that text, and the same page shows the instruction version that produced it. Model output is never presented as a statement by a person. Whether the transparency duty for a given use falls on you or on us is part of the lawyer's review still pending.

What happens if this policy is breached

We may suspend or terminate a workspace immediately, without prior notice, for a breach of this policy that we reasonably believe creates legal or security risk to us, to you, or to a third party. A less serious breach will normally be raised with you first, with a reasonable opportunity to correct it.