An agent that answers on your behalf can reach a great many people quickly. This policy sets out what it may not be used for. It applies to every deployment, pilots included, and forms part of the Terms of Service.
Effective 1 August 2026
The agent must not claim to be a specific named human, and must not deny being an automated system when a person sincerely asks. It may have a name and a personality. It may not impersonate a real individual, and it may not be configured to assert that it is human.
The service may not be used to commit, plan or facilitate anything unlawful under the laws that apply to you or to the people it speaks with. That includes fraud, money laundering, sanctions evasion, and the sale of goods or services you are not licensed to sell.
No use for harassment, intimidation, stalking, threats, or the targeting of an individual with unwanted contact. No use to demean people on the basis of race, religion, nationality, gender, sexuality, disability or any comparable characteristic.
The agent may not be pointed at purchased lists, scraped contacts or anyone who has not agreed to hear from you. Bulk unsolicited calling or messaging is prohibited regardless of what local law happens to permit.
The agent may not deliver medical diagnoses, prescriptions, legal advice, or individual financial or investment recommendations as though it were qualified to do so. It may take details, answer factual questions and book appointments; it must hand over to a qualified person for the judgement itself.
No use to make final decisions on credit, insurance, housing, employment, education or benefits, and no use as the only channel for emergencies. An agent may gather and prepare. A person decides, and a person must always be reachable.
Do not configure the agent to solicit health conditions, biometric data, religious or political affiliation, sexual orientation, or precise location unless you have a lawful basis and have told us in writing so the deployment can be handled appropriately.
The service is not intended for use with children, and must not be configured to market to them or to collect their personal data.
No attempts to extract the underlying models or system instructions, to bypass configured limits, to overwhelm the platform, or to use it as a route into systems you are not authorised to reach.
Report suspected misuse to contact@orikin.ai. Security vulnerabilities should follow the disclosure process at /.well-known/security.txt.
We would rather correct a configuration than close an account, and in most cases a conversation resolves it. Where it does not, we may suspend a deployment. Where there is a risk to people, we may suspend it immediately and without notice.