Acceptable Use Policy

Scope

This policy applies to every use of Hania — production, dev, internal testing, or research. It applies to you, every member of your workspace, every API consumer using your keys, and every end user interacting with an agent you've built. You're responsible for all of them complying with this policy.

Prohibited use cases

You may not use Hania, or build agents on Hania, to:

• Generate or distribute illegal content — including CSAM, content that incites violence, terrorism content, or anything illegal in the user's jurisdiction.
• Defraud, deceive, or impersonate — agents must not pretend to be a specific real human (yourself, a public figure, or a customer) without their explicit consent. Brand-voiced personas are fine; impersonation is not.
• Operate phishing, scam, or social-engineering campaigns — including pretexting calls, voice phishing ("vishing"), or fake "tech support" outreach.
• Practice medicine, law, or other regulated professions without appropriate licensing and human supervision. Agents may provide general information; they may not give individualized professional advice as the responsible party.
• Make automated decisions with legal or similarly significant effects on people (employment, credit, housing, insurance) without meaningful human review.
• Operate political campaigns or voter outreach at scale, including robocalls and synthetic political ads.
• Generate non-consensual intimate imagery, sexual content involving minors, or content that sexualizes real people without their consent.
• Surveil people without their knowledge or against their consent — including covert recording of phone calls in jurisdictions where two-party consent is required.
• Build weapons — biological, chemical, nuclear, radiological — or to provide meaningful uplift to anyone trying to.

Prohibited techniques

• Don't bypass our safety systems. No jailbreaks, prompt injection scaffolding, or running our models in ways that intentionally circumvent content policies.
• Don't crawl us. The web dashboard isn't an API. Don't scrape it. Use the official API or SDKs.
• Don't proxy us by re-selling Hania access in a wrapper that pretends to be a different service. (Building products on top of Hania is fine — re-selling raw model access dressed as your own model is not.)
• Don't load-test in production without telling us first. Synthetic loads are fine on staging; talk to us about prod ramps.

AI disclosure

End users must know they're talking to an AI when they reasonably need to. Specifically:

• Voice agents must identify themselves as AI in the opening turn, or in response to any direct question about whether they are human.
• Chat widgets must show an AI label in the launcher and again at the start of every session.
• You may not coach the agent to deny being an AI when a user asks.

End-user safety

You must build appropriate handoff and safety paths:

• Self-harm or crisis — if your agent will be exposed to consumers, you must wire in a crisis-response handler that surfaces local hotlines and routes to a human where available.
• Emergencies — agents may not be the only path to reach you in an emergency. Always provide an alternative.
• Vulnerable populations — extra care for agents that talk to children, elderly users, or users in medical or financial distress.

Abuse & security

Don't try to break Hania. Specifically: no attempts to compromise other tenants' data, no DoS attacks, no reverse-engineering our models. Coordinated disclosure for vulnerabilities is welcomed — see our bug bounty program.

Enforcement

When we believe you've violated this policy, our response depends on the severity:

• Minor or accidental — we'll reach out and ask you to fix it. Plenty of time, no penalty.
• Material violation — we may suspend the violating agent or workspace pending remediation.
• Egregious or unlawful — we will terminate immediately, preserve evidence, and where required, notify law enforcement.

We try to be fair and proportionate. We also try to be fast — abuse hurts everyone.

Report a violation

Seen something? Email [email protected] with as much detail as you can. We review every report within one business day. Anonymous reports are accepted.