Acceptable use policy

• Legitimate work product subject to professional privilege (legal, medical, accounting, journalism).
• Research, analysis, and drafting on confidential matters that your home jurisdiction permits you to do.
• Routine engineering work where IP / contractual confidentiality applies.
• Any use case mainstream AI providers serve, where the buyer’s preference is to architecturally limit the operator’s exposure to the contents.

• Generation of content that constitutes a crime in Switzerland (e.g. CSAM, materials inciting violence against specific persons, certain categories of fraud).
• Active assistance with the planning or execution of physical violence, terrorism, or weapons of mass destruction.
• Evasion of sanctions imposed by the Swiss government or applicable international bodies.
• Targeted harassment campaigns or doxxing against named individuals.

This list is not exhaustive. We picked the categories that matter to the buyer segment we serve. We will not be the “anything goes” AI provider.

We do not see the contents of your dispatches in the confidential cloud. Enforcement happens at three observable boundaries: (1) metadata patterns that strongly correlate with prohibited categories (e.g. extreme volume against a specific named target), (2) credible third-party reports, and (3) on the open-source dispatcher, ledger telemetry an operator chooses to act on. We terminate tenants on credible evidence. We do not act on speculation.

Suspected AUP violations should be reported to abuse@xanadu.run with as much detail as possible. We will respond within 5 business days and post a summary of enforcement actions in our quarterly transparency report (without naming customers).