Terms of Use

1. What Zaevi is

Zaevi is a consumer AI tool that classifies your query and routes it to the most appropriate AI model. Outputs are generated by third-party AI models (from Anthropic, Google, Meta, Mistral, DeepSeek, Microsoft, Qwen, and others). Most models are reached through OpenRouter; Anthropic's Claude models are called directly on Anthropic's API. Zaevi does not train its own model.

2. Eligibility & conditions of use

• You must be at least 16 years old to use Zaevi.
• One human, one account. No automated scraping, framing, or re-hosting Zaevi's output as your own product without prior written permission.
• Don't hammer the system. Reasonable rate limits apply; we may throttle or block excessive requests.
• We may change these terms. Material changes will surface as a visible notice on the site. Continued use after a change means you accept the change.

3. What you must NOT submit

You agree not to use Zaevi to:

• Submit content that is illegal where you live or where Zaevi operates.
• Generate child sexual abuse material, non-consensual intimate content, content designed to harass or defame a specific person, or content that promotes violence against a person or group.
• Attempt to extract someone else's private information, credentials, or to break into systems.
• Generate disinformation designed to deceive at scale (fake news campaigns, impersonation of real people for political manipulation, fabricated quotes from public figures).
• Probe Zaevi's safety systems with the intent to weaponize outputs (jailbreak research is fine; productionizing the result against real users is not).

You are responsible for what you submit and what you do with the output. If you misuse Zaevi and a third party comes after us because of your actions, you indemnify us.

Location sharing is optional and opt-in. If you grant location access, you can revoke it anytime in your browser or phone settings, see §5 for the exact path.

4. AI-generated content: not professional advice

Zaevi's outputs are not medical, legal, financial, mental-health, tax, immigration, or any other kind of professional advice. They are conversational answers from an AI model that may be confidently wrong.

How Zaevi works under the hood. Every query is classified and routed to the smallest model that can produce a competent response. Most queries, including those touching on consequential life decisions (career changes, relationships, healing, financial moves, identity work), are served by smaller, less computationally intensive models. Output quality varies by model, by topic, and by how much context you give Zaevi to work with.

What this means for you:

• Do NOT rely on Zaevi's output for any decision with real-world consequences without independent verification from qualified sources.
• For decisions that carry real consequences, consult a licensed professional: doctor, lawyer, accountant, independent financial advisor, therapist. Zaevi is not a substitute for any of them.
• Zaevi may sound certain. Treat that certainty as conversation, not as fact. Outputs are AI-generated and can be inaccurate, outdated, or wrong in ways that aren't obvious.
• If Zaevi's response turns out to be wrong and you act on it, the responsibility for any resulting harm rests with you, not with Zaevi or its operators (see §6).
• If you are in a mental-health crisis or in immediate danger, call your local emergency number — 112 across the EU (911 in North America) — or a crisis line: in the Netherlands, 113 Zelfmoordpreventie (call 113 or 0800-0113, free, 24/7); 116 123, which works in the Netherlands and a number of other European countries; in the US, 988. Zaevi is not a crisis service.

5. Data & logging

We keep a record of every turn — which model answered, how long it took, what it cost, and the routing metadata behind it. That record is what we use to improve the classifier and debug failures, and none of it is the words you typed.

Whether we store the words is a separate question, and if you have an account it is your choice. Settings has a "Saving conversations" switch. Leave it off and we write the technical record and nothing else — your text is never stored, which also means Recents stays empty. Turn it on and your conversations are kept for 30 days, then the text is erased automatically. The technical record stays either way.

Without an account there is nothing to attach that choice to, so conversations are stored and the text is erased on the same 30-day schedule.

We run an automated PII redaction pass on stored queries and responses (emails, phone numbers, credit cards, government IDs, IPs, URL tokens) before write, but redaction is not perfect. Don't submit anything you wouldn't put in a private email to a colleague. We do not sell your data, do not use it to train models, and do not share it with advertisers. The third- party model providers we route to have their own logging policies. On OpenRouter we set the "model training" toggle to OFF, which prefers no-log providers where possible; Anthropic's Claude models, which we call directly, are governed by Anthropic's commercial API terms — your inputs are not used for training, with limited retention for abuse monitoring only.

Location. Sharing your location is opt-in. If you grant access, your browser reads your coordinates and sends them straight to a third-party geocoding service (BigDataCloud) to turn them into a place name. That request leaves your device, not ours: your raw coordinates never reach Zaevi's servers at any point. What reaches us is only the city / neighborhood label (e.g. "De Pijp, Amsterdam"), and only when you send a query while location is enabled. The label is sent with that query to help ground recommendations geographically; it is not stored. You can revoke location access anytime in: Settings → Site Settings → Location on Android, Settings → Safari → Location on iOS, browser settings on desktop.

You can delete your account and everything tied to it from Settings at any time. You can delete a single conversation, or all of them, from Recents. And you can download everything we hold about you from Settings — you do not have to ask us for any of this. If you would rather we did it, or you want something these controls do not cover, email privacy@zaevi.app. We aim to respond within one month.

6. No warranty, limited liability, decisions are yours

Zaevi is provided as-is. We make no guarantee that outputs are accurate, complete, current, or fit for any particular purpose.

Decisions you make based on Zaevi's responses are yours. Whether a recommendation works out, fails, or causes harm depends on factors Zaevi cannot see: your full life context, the specific people involved, your local laws, your medical history, your finances, your timing. Zaevi only ever offers a perspective shaped by its system prompt and the model serving the query. The decision to act, and the consequences of acting, are yours alone.

To the maximum extent permitted by applicable law, Zaevi and its operators are not liable for any direct, indirect, incidental, consequential, or punitive damages arising from use of the service, including any decision made on the basis of an output. Use it at your own risk.

7. Termination

We may suspend or terminate access for any account at any time if we believe these terms have been violated. You can stop using Zaevi at any time.

8. Contact

Questions about these terms, or any request about your data — a copy of it, a correction, or deletion — privacy@zaevi.app.