Shardio Privacy Policy
Last updated: 2026-10-26
This Privacy Policy explains how Shardio, LLC ("Shardio", "we", "us") collects, uses, discloses, and protects personal information in connection with the Shardio websites, console, APIs, node software, and related services (the "Services"). Shardio, LLC is the controller of the personal information this policy describes. Where a business uses the Services to process personal data it controls, we process that data on the business's behalf under our Data Processing Addendum.
1. Scope
This policy applies when you visit our websites or console, join the waitlist, create or use an account or API key, submit prompts or other inputs to the Services, run node software as a host, or contact us.
2. What We Collect
Account and identity. Your name, email address, billing and notification email if you set them, and a subject identifier from our identity provider; API keys and account identifiers. We never hold passwords for the Services. If you join the waitlist, your email address.
Payment and billing. Through our payment processor we keep a customer identifier and, where you turn on automatic top-ups or authorize automatic payment of invoices, a payment-method reference. We do not receive full card numbers. If we bill your account after use (postpaid billing), the invoicing details you give us: legal name, billing address, billing email, tax identifier, and a purchase order number if you provide one. If you withdraw earnings, we keep the identifier of your payout account with our payment processor, which collects the bank and tax details it requires. Balances, credit history, usage charges, invoices and their payment status, earnings, transfers, withdrawals, and statements.
Usage and service metadata. Request identifiers, timestamps, selected model, token counts, timings, outcome, cost, and which node served a request. IP addresses in server logs, security records, and device sign-ins.
Content (Inputs and Outputs). Prompts, messages, images, and other inputs you submit, and the completions returned. Section 4 explains how content is handled.
Host device data. Node and machine identifiers, hardware class and capacity, public IP, heartbeats and uptime, platform check and calibration results, software versions and image digests, and the account each machine belongs to.
Network and connection information. We collect IP addresses and other connection information to operate and secure the Services, connect machines, deliver requests, diagnose problems, and prevent abuse. We also use this information to estimate the country, region, and network from which a serving machine connects. These estimates are not verified physical locations. Our location records retain the derived information and observation time, not the source IP address; IP addresses may separately appear in operational and security logs.
Website. On shardio.ai, the campaign information in the link you followed, and the advertising measurement described in Section 8.
Support. Messages you send us and our replies.
3. How We Use It
To provide and secure the Services; authenticate accounts; route requests and return outputs; meter usage and bill accurately; pay hosts; detect fraud and abuse; enforce our Terms; meet legal obligations; measure our advertising; and communicate about the Services (including balance warnings, receipts, and invoices). We use aggregated usage metadata for reliability analytics. We do not train models on your Content, and neither this policy nor our Terms permits it.
4. How We Handle Content
Because inference is the product, content handling deserves plain language:
- Our gateway processes your requests. It reads them to route, meter, and bill them, and holds content in memory only while it serves a request. Our request logs record metadata, not prompt or completion content.
- Our billing ledger stores metadata, not content. Each request leaves a record of token counts, timings, model, outcome, and cost, never the prompt or completion itself. We do not store request content.
- Hosts run our software on their hardware. Your request is served by a model engine on hardware an independent host operates. Shardio's node software passes the request to the engine in an isolated execution process inside the container image we publish, whose signature our installer and CLI check before it starts, and it runs the engine without request logging. Hosts agree in our Terms and our Host Data Processing Addendum not to inspect, retain, or misuse content that passes through their hardware. Hosts operate the hardware, so these protections are contractual and operational rather than hardware-based confidential computing.
- Engines cache recent prompts in memory. To answer faster, the engine keeps a cache of recently processed prompt text in the serving machine's memory. The cache is partitioned by account, so it is reused only for later requests from the same account, and it is overwritten as the engine serves new requests. Our current serving configurations keep no such cache on disk.
- A request may be completed by another node. If a node fails before it starts answering, our gateway may send the same request to another eligible node to complete it.
- Checks and calibration use our own requests. Platform checks and calibration use requests Shardio writes, never drawn from members' prompts, and responses to them may be compared across nodes. We do not sample or replay members' requests to check nodes.
- Playground conversations stay in your browser. The console's Playground sends each turn with the conversation so far; the conversation itself is kept in your browser tab, not on our servers.
- Distribution partners. If you reach us through a distribution partner, that partner handles your request before it reaches us, under its own privacy policy.
5. Service Providers and Other Recipients
We share personal information with service providers that run parts of the Services for us:
| Provider | What it does for us |
|---|---|
| Clerk | Sign-in and the waitlist |
| Stripe | Payments and payouts |
| Resend | Account emails |
| Fly.io | Hosting and database |
| Tigris | Storage |
| Upstash | Live balances and rate limits |
| Cloudflare | The shardio.ai website and DNS |
| X | Advertising measurement on shardio.ai only, where enabled (Section 8) |
We also share information with hosts, to the extent a request is processed on their hardware (Section 4); hosts are bound by our Host Data Processing Addendum; with law enforcement or regulators where required by law; in connection with a merger, acquisition, or financing; and with your direction or consent. We do not sell personal information.
6. Retention
| Data | Retention |
|---|---|
| Request content (prompts, completions) | Not stored; held in memory while a request is served, and in engine caches as described in Section 4 |
| Request records (token counts, timings, model, outcome, cost; no content) | Kept for 180 days, then deleted in monthly batches once their totals are recorded in your account's billing history |
| Server logs | Kept for 90 days |
| Node uptime history | Kept for 90 days |
| Account and identity data | Deleted 30 days after the account is closed, or after outstanding invoices are resolved if later |
| Billing and payout records (ledger, credit history, statements, invoices, earnings, withdrawals) | Kept while your account is open, and for 7 years after your last payment or payout once it is closed, for tax and accounting |
| Host platform check and calibration history | Kept for 2 years after a node leaves, for trust and dispute purposes |
| Support messages and incident notes | Kept for 2 years after the conversation closes, then deleted |
| Backups | Backups are encrypted and stored off-site. Daily backups are kept for 35 days and monthly backups for 12 months. Deleted data may remain in a backup until it rotates out. |
7. Your Rights
Depending on where you live you may have rights to access, correct, delete, or port your personal information, to object to or restrict certain processing, to withdraw consent you have given, and to opt out of certain processing. We may ask for information needed to verify a request where the law permits, and may retain what accounting or legal requirements demand (billing and payout records in particular). Requests: support@shardio.ai. If a business sent us your information through its use of the Services, please also contact that business, which decides how your information is used; we will help it respond.
If you are in the European Economic Area or the United Kingdom. We rely on these legal bases:
| Processing | Legal basis |
|---|---|
| Providing the Services, including processing your requests, your account, and payments | Performance of our contract with you |
| Billing, tax, and accounting records | Compliance with legal obligations |
| Security, fraud and abuse prevention, platform checks, and reliability analytics | Our legitimate interests in running a safe and reliable service |
| Service emails such as receipts, invoices, and balance warnings | Performance of our contract with you |
| Advertising measurement on shardio.ai | Your consent where the law requires it, otherwise our legitimate interest in measuring our advertising |
You may complain to the data protection authority where you live or work, or where you believe an infringement happened. In the United Kingdom, that is the Information Commissioner's Office. We would appreciate the chance to address your concern first at support@shardio.ai.
8. Cookies, Browser Storage, and Advertising
Our identity provider uses cookies to keep you signed in, and the console stores your preferences in your browser. On shardio.ai, your browser session keeps the campaign information in the link you followed so we can tell which campaign brought you. Where the X pixel is enabled on shardio.ai, it measures site visits and waitlist sign-ups for our advertising, together with that campaign information. We use it for advertising measurement only: X's restricted data use setting is on, and we do not use it to build advertising audiences or to target ads. The console and API use no advertising trackers.
9. Security
Connections to our website, console, and API use HTTPS. Connections between our gateway and hosts' nodes are encrypted, and our own services communicate over an encrypted private network. Our gateway and the serving engine process request content in readable form in order to serve it (Section 4). Backups are encrypted. Secrets are generated, never committed to code.
10. International Transfers
The Services are available worldwide. Our infrastructure is based in the United States, our service providers may process information in other countries, and hosts may operate their hardware in any country. Your request may be served by a host in a country other than your own; we do not route requests by region.
When we transfer personal information from the European Economic Area, the United Kingdom, or Switzerland to another country, we rely on an adequacy decision where one covers the recipient, and otherwise on the European Commission's Standard Contractual Clauses, with the UK International Data Transfer Addendum for UK transfers and the adjustments Swiss law requires for Swiss transfers. Hosts are bound by those clauses through our Host Data Processing Addendum. You may ask for a copy of the safeguards at support@shardio.ai.
11. Children; Changes; Contact
The Services are not directed to children under 13, and users must be 18+ (Terms §1). We will email you at least 30 days before a material change to this policy takes effect, except where a change is required sooner for legal or security reasons, and every version carries its effective date. Contact: support@shardio.ai.