Acceptable Use Policy
This Policy applies to every way of working with the Service: the API, the web chat, the Telegram bot and any integration that uses your API key. It applies to you and to your employees, contractors and the end users of your applications: you are answerable to the Operator for their actions.
1. Express prohibitions
The Service may not be used to:
- break the law or infringe other people’s rights, or to call for, assist or incite such conduct;
- create, seek out or distribute child sexual abuse material or any material that exploits children — such requests lead to immediate blocking without warning and, where required by law, to notification of the competent authorities;
- create intimate imagery of people without their consent;
- plan, support or promote terrorism, violence or harm to people, or to develop weapons, including biological, chemical, nuclear and radiological weapons;
- create malware, exploits or tools for unauthorised access to other people’s systems, or to carry out attacks, scanning or credential stuffing;
- engage in fraud, phishing, pyramid schemes, forged documents, circumvention of identity verification, or impersonation of the voice or appearance of real people in a misleading way;
- send spam and unsolicited bulk messages, including generating the text for such campaigns;
- harass, stalk or degrade people, or incite hatred on grounds such as race, nationality, religion, sex or disability;
- produce disinformation, pass generated material off as official statements of people, organisations or public authorities, or interfere with elections;
- make decisions without human involvement that materially affect people’s rights — on lending, hiring, dismissal, insurance, medical care, education, or immigration or legal status;
- process biometric data, health data, criminal records or other special categories of personal data without a lawful basis, or use the Service for de-anonymisation and surveillance;
- infringe intellectual property rights: circumvent technical protection measures, reproduce other people’s works as your own, or extract protected content from third-party databases.
2. Service-related restrictions
- API keys may not be transferred, sold or shared with third parties, and access to the Service may not be resold as your own offering without the Operator’s written consent.
- Rate limits, quotas, charging and safety filters may not be circumvented, including by registering multiple accounts for that purpose.
- The Operator’s infrastructure and other clients’ use of it may not be disrupted, and the Service may not be probed for vulnerabilities without prior written permission. Report any vulnerability you find to hello@cloudai.am — we respond to such reports and do not pursue good-faith disclosure.
- Model output may not be used to train competing models where the rules of the relevant model provider prohibit it.
- In applications for end users it must be clear that the interlocutor is software rather than a human, where the contrary could mislead the user.
3. Model providers’ rules
A request sent through the Service is executed by the model’s developer. Their rules therefore apply to you directly, and breaching them automatically breaches this Policy — even where the specific prohibition is not listed in section 1.
The current rules are published by the providers themselves, in particular:
| OpenAI | Usage Policies — openai.com/policies/usage-policies |
|---|---|
| Anthropic | Usage Policy — anthropic.com/legal/aup |
| Generative AI Prohibited Use Policy — policies.google.com/terms/generative-ai/use-policy | |
| xAI, Meta, Mistral, DeepSeek | the acceptable use policies of the respective developers, published on their official websites |
By choosing a model you accept the rules of its developer. Where a provider’s rules are stricter than this Policy, those rules apply.
4. Eligibility and territorial restrictions
The Client is solely responsible for being entitled to use the Service and the selected models: for not being subject to sanctions, not acting in the interests of a sanctioned person, and for access to the relevant model not being restricted for their jurisdiction by the provider’s rules or by applicable export control law.
Using the Service to circumvent a model provider’s geographic restrictions is prohibited. The Operator does not provide access in circumvention of such restrictions and may refuse service or block access where a provider, a payment system or a competent authority so requires.
5. How we enforce this Policy
The Operator does not read the content of your requests or responses — this is stated expressly in clause 5 of the Terms of Use and is not altered by this Policy. Enforcement therefore relies on what is available without access to content:
- metadata and usage anomalies — sharp spikes, unusual request patterns, signs of automated abuse;
- reports from affected persons and rights holders;
- notices and claims from model providers, who do inspect content on their side;
- requests from competent authorities made in accordance with the law.
Where investigating a specific incident requires access to content, this is possible only with your consent or pursuant to a lawful request from a competent authority.
To report a violation: hello@cloudai.am. Describe what happened and attach evidence where possible. We review such reports within 3 business days.
6. Consequences of a breach
- A warning with a request to remedy the breach — for minor and remediable breaches.
- Restriction or suspension of access — where the breach continues, where there is a risk to the infrastructure or to other clients, or where a model provider so requires.
- Immediate blocking without warning — for material that exploits children, for preparation of violence, for attacks on the infrastructure, and in other cases where delay causes harm.
- Termination of the contract and refusal of further service.
Treatment of the remaining balance upon blocking for a breach: [procedure not yet approved, see clause 4 of the Terms]. The Operator may claim compensation for losses, including amounts claimed by providers and third parties.
7. Changes to this Policy
This Policy is updated as legislation and providers’ rules change. A new version is published on this page with its date; material changes take effect 14 days after publication.