This policy explains what ARC Solutions Inc ("ARC", "we", "us") does with information when you use the ARC Kimi K3 API, directly or through a platform such as OpenRouter. ARC Solutions Inc is located in Wilmington, Delaware 19803, United States.
1. Our commitments
- Zero data retention. We do not store, log or keep the content of any live request or response. Batch files, where we offer batch processing, are deleted 24 hours after the batch completes, is cancelled, expires or fails.
- No training. We do not use requests or responses to train, fine-tune or evaluate any model.
- United States. We serve and process requests in the United States.
- Metadata only. What we keep about a request is operational metadata for billing and reliability, never its content.
2. Data you send to the API and the responses it returns
Live requests to the API (requests answered one at a time, as opposed to batch jobs) can contain prompts, messages, system instructions, images, tool definitions, tool results and structured-output schemas. Responses contain generated text, the model's reasoning and tool calls. We do not store, log or keep the content of any live request or response: it is not written to disk, and it is not kept in logs or backups. It is not kept for debugging, quality review, abuse detection or safety review.
While a request runs, its content is held only in the working memory of the servers that answer it. When the answer is complete, nothing of it is kept except derived data in a cache: to avoid computing the same input twice, the model servers keep the model's intermediate results for recent prompts, and the processed form of recent images, in their own memory. That cache is never written to disk, is used only to answer later requests that begin with exactly the same input, is evicted as space is needed, and is cleared whenever a server stops. Our model servers run for at most 24 hours before they are replaced, so nothing in the cache lasts longer than 24 hours. Images sent as links are fetched by the model server when the request runs and are not stored.
We do not use requests or responses to train, fine-tune or evaluate any model. We do not sell them, and we do not share them, except that the infrastructure providers listed by category in section 6 process them for us, under our instructions, to answer them.
3. Batch processing
When we offer batch processing, a batch's input and output files must be held until the batch ends so that you can collect the results. They are deleted 24 hours after the batch completes, is cancelled, expires or fails, together with the text of every request in the batch. A file you upload that is never used in a batch is deleted 24 hours after upload, and you can delete any file sooner through the API. Batch files are never used for training.
4. Information we collect
- Operational metadata for each request: a request id; which API key was used (its id, never the key itself); the model; token counts (input, cached input, output and reasoning); timing (time to the first token and the total); the outcome (completed, cancelled by the caller, or the kind of error); and which of our servers answered. It contains no request or response content.
- Account information for direct customers: the email address and other details you give us when you create an account, your password and API keys (each stored only as a one-way hash), the network (IP) address you sign up from, your credit balance, including any free credit, and your billing history. Card and payment details are collected and held by our payment processor, not by us.
- Free credit limits: to limit free credit per account, per browser and per network, we keep a keyed hash of your network address and a hash of a cookie that identifies your browser.
- Cookies: a sign-in cookie and the browser cookie above. We use no advertising or analytics cookies.
- Messages you send us, for example to contact@arc.ai.
Our infrastructure providers may process network information, such as IP addresses, to deliver requests to us and to protect their services.
5. How we use information
We use the information above to provide and secure the API, to bill for usage and keep accounts, to plan capacity and keep the service reliable, to answer your messages, and to meet our legal obligations.
6. How we share information
We do not sell personal information. We share information only as follows:
- Cloud infrastructure providers in the United States, which process requests in memory under our instructions and receive no rights to the data.
- Payment processing, for direct customers who buy prepaid credit.
- Platforms you use to reach us, such as OpenRouter, which send us your requests and receive our responses under their own policies.
- When the law requires it, or to protect the rights, property or safety of ARC, our customers or others.
- In a merger, acquisition or sale of assets, under a policy that protects information at least as well as this one.
A current list of our sub-processors is available on request at contact@arc.ai.
7. Where requests are processed
United States. We serve and process API requests on servers in the United States.
8. Data retention
- Live request and response content: not retained (section 2).
- Batch files and the text of batch requests: deleted 24 hours after the batch completes, is cancelled, expires or fails (section 3).
- Operational metadata: kept for as long as we need it for billing, accounting and security, and to meet legal obligations.
- Account information: kept while your account is open, and afterwards as long as needed to meet legal obligations.
9. Security
Requests and responses travel over encrypted connections. Passwords and API keys are stored only as one-way hashes, access to our systems is limited to the people who run them, and the content of live requests is never written to disk.
10. Your choices and rights
Because we keep no live request or response content, there is none to access or delete; batch files can be deleted through the API at any time. For account information, you can ask us to access, correct, export or delete it by writing to contact@arc.ai. Depending on where you live, you may have further rights under local law, and we will honour them. We will not treat you differently for using them.
11. Children
The API is not directed to children under 18, and we do not knowingly collect personal information from them.
12. Changes to this policy
If we change this policy, we will post the new version here and update the "Last updated" date before it takes effect.
13. Contact
ARC Solutions Inc, Wilmington, Delaware 19803, United States. Email: contact@arc.ai.