Chutes Global Corp., a Delaware corporation ("Chutes," "we," "us," or "our"), provides this Privacy Policy to explain how we collect, use, disclose, retain, and protect personal data when you use chutes.ai, our application programming interfaces (including api.chutes.ai and llm.chutes.ai), our developer tools, Chutes Chat, Chutes Search, our confidential computing products, and related services (the "Platform"). Capitalized terms that this Privacy Policy does not define have the meanings given in our Terms of Service.
If your organization has signed a Data Processing Agreement with Chutes, or uses the Platform under our Data Processing Agreement, that agreement governs our processing of personal data in your Inputs and Outputs, and it controls over this Privacy Policy where the two conflict.
Our core commitments
- Chutes does not store your API content. Chutes does not log, store, or persist the content of your API requests or responses on Chutes systems, subject only to the limited exceptions in Section 6.3. In end-to-end encryption mode, our application layer cannot read your content at all.
- Chutes does not train on your content. We do not use your Inputs or Outputs to train, fine tune, distill, evaluate, or otherwise improve any model.
- We do not sell personal data.
- We tell you where your request runs. Every model page shows whether a model runs in a trusted execution environment ("TEE"), and our model catalog reports the same fact in the
confidential_computefield. - We tell you what third parties may keep. When a third party inference platform serves your request, Section 4 explains what that platform may retain and for how long.
1. Who we are and how to reach us
Chutes Global Corp. is the controller of the personal data described in this Privacy Policy, except where Section 9 explains that we act as a processor or service provider for a business customer.
Chutes Global Corp. 2810 N Church St., Suite 89723 Wilmington, Delaware 19802 United States Privacy and legal requests: legal@chutes.ai General support: support@chutes.ai
2. How your request is served
The privacy protection that applies to a request depends on the type of Model that serves it. There are two types.
2.1 Confidential Compute Models (TEE)
A Confidential Compute Model is a Model that our catalog reports with confidential_compute: true and that the model page labels as TEE. These Models run on compute supplied by independent operators on the Chutes subnet of the Bittensor network ("Network Operators"), inside trusted execution environments that use Intel Trust Domain Extensions and NVIDIA confidential computing with Protected PCIe. Before a TEE workload starts, our validator checks hardware attestation evidence against a known good configuration.
We designed this configuration so that the Network Operator that physically controls the hardware cannot read your content while it is processed. If you also use our end-to-end encryption mode, your content is encrypted on your device, decrypted only inside a verified TEE, and encrypted again before it leaves, so the Chutes application layer relays only ciphertext and a plaintext usage summary for billing.
Hardware isolation reduces risk but does not eliminate it. Trusted execution environments remain subject to processor, firmware, and side channel vulnerabilities and depend on the integrity of the attestation chain. We monitor vendor advisories, apply security updates as they become available, and will notify affected customers of any vulnerability that we determine materially affects these protections.
2.2 Standard Models
A Standard Model is any Model that is not a Confidential Compute Model. A Standard Model does not run inside a Chutes TEE.
Many Standard Models are served by established third party inference platforms ("Model Providers"), which run the Model on their own infrastructure. When you use one of these Models, Chutes sends your Inputs to the Model Provider, and the Model Provider returns the Outputs through us. Model Providers may operate infrastructure in the United States and in other countries, including countries in Asia. Chutes does not operate a Model Provider's systems, and a Model Provider's own staff and systems can access content while they process it.
Because a model page shows only whether a Model runs in a TEE, you should assume that any Standard Model may be served by a Model Provider. If a particular Model Provider's terms require us to identify it to you, we will identify it on the relevant model page.
2.3 Choosing the right Model
If your content is confidential or sensitive, use a Confidential Compute Model, and use end-to-end encryption where it is available. Our Terms of Service prohibit submitting certain categories of sensitive data to Standard Models. If you use an automatic or fallback routing option, the Platform may select a Standard Model, so request a Confidential Compute Model by its identifier if you need TEE protection for every request.
3. Personal data we collect
3.1 Data you give us
- Account data. Your email address, username, account identifier, authentication credentials or sign in assertions from an identity provider you choose, and API keys. For business accounts, your organization's name and your role.
- Billing data. Billing contact details, transaction records, invoices, balances, and plan information. Our payment processor, Stripe, handles full payment card numbers, and we receive only limited card details such as the last four digits and expiration date. If you pay with cryptocurrency, we process your wallet address and on chain transaction identifiers.
- Verification and compliance data. Where the law requires it for sanctions screening, fraud prevention, or tax purposes, identity and organization information such as legal name, address, and ownership information.
- Inputs and Outputs. The prompts, files, images, audio, video, and other content you submit, and the content the Platform returns. Sections 4 and 6 explain how we and Model Providers handle this content.
- Support and deployment data. Support tickets and correspondence, and the definitions, images, configuration, and repository references you submit to deploy a User Chute.
3.2 Data we collect automatically
- Usage metadata. Timestamps, request durations, token counts, model identifiers, error codes, latency, throughput, and the Model and instance that served a request. We use this metadata for billing, capacity planning, reliability, and security. It does not include the content of your requests.
- Device and connection data. Your IP address, the approximate location we derive from it, browser type, operating system, language, and time zone.
- Short lived operational values. Rate limit counters, session tokens, and one way hashes of prompt prefixes that we compute for short lived request routing and caching. Where we can associate a hash with an account, we treat it as personal data and keep it only for the short period Section 7 states.
- Cookies and similar technologies. Section 8 describes these.
3.3 Data from third parties
We receive transaction outcomes and fraud signals from payment processors, screening results from sanctions and fraud prevention vendors, identity assertions from sign in providers you choose, security signals from our content delivery and network protection providers, and hardware attestation and performance data from Network Operators, which does not include your content.
3.4 IP addresses
We do not store IP addresses in our core product databases alongside request content. We use IP addresses and approximate location for rate limiting, abuse and fraud prevention, security, and sanctions screening, and we keep them in security logs for the period Section 7 states. We do not send your IP address to Network Operators or Model Providers as part of routing your request, although a Model Provider may see network information about Chutes' own servers.
4. Standard Models served by Model Providers
This Section explains what happens to your content when a Model Provider serves a Standard Model. It reflects the agreements and published policies that apply to the Model Providers serving the Platform on the date shown above.
4.1 What Model Providers may retain
Depending on the Model Provider and the Model, a Model Provider may hold your content as follows:
| Category | What a Model Provider may keep | How long |
|---|---|---|
| Zero retention processing | Where a Model Provider offers zero data retention for a Model, your Inputs and Outputs are not written to durable storage. They may pass through an encrypted cache needed to complete the request. | Deleted when the request completes. Any cache is automatically deleted, generally within 24 hours. |
| Standard provider retention | Where zero data retention does not cover a particular Model or feature, the Model Provider's standard retention applies to your Inputs and Outputs. | Generally up to 30 days, or longer where the law or the security of the Model Provider's service requires. |
| Files and multistep features | Files, images, and other inputs submitted to features that process a request in several steps or asynchronously. | Until the task completes and, under zero retention, generally no longer than 48 hours. Otherwise under the Model Provider's standard retention. |
| Content flagged by safety systems | Inputs and Outputs that a Model Provider's automated safety systems flag as potentially violating the law or its policies. | Up to two years, for safety review, abuse prevention, and legal compliance. |
| Legal holds and legal requests | Content subject to a legal obligation, court order, or valid government request directed to the Model Provider. | For as long as the obligation requires. |
| Request metadata | Timestamps, token counts, model identifiers, request identifiers, and error codes. | For the Model Provider's billing, security, and service level purposes. |
4.2 Training
Our agreements with the Model Providers that serve Standard Models do not permit them to use your Inputs or Outputs to train, fine tune, or improve their models without our separate consent, and we do not give that consent.
4.3 Safety and moderation
Model Providers may run automated safety, abuse detection, and content moderation systems on the content they process. Those systems may block a request, filter an Output, or flag content for retention as Section 4.1 describes.
4.4 Location
Model Providers may process your content in the United States and in other countries, including countries in Asia. Section 13 explains how we handle international transfers.
4.5 Our limits
These are commitments that Model Providers make to us by contract or in their published policies. Chutes does not operate Model Providers' systems and cannot guarantee their compliance. We review Model Providers before we route traffic to them, and we act on credible reports of noncompliance, including by suspending routing. If the retention commitments of a Model Provider become less protective than Section 4.1 describes, we will update this Privacy Policy before we route your content to that Model Provider under the new terms.
4.6 Information for business customers
A list of the Model Providers that serve Standard Models, with their processing regions, is available on request to customers that have a Data Processing Agreement with Chutes. Write to legal@chutes.ai.
5. How we use personal data, and our legal bases
We use personal data for the purposes below. If you are in the European Economic Area, the United Kingdom, or Switzerland, the table also shows the legal basis we rely on under the General Data Protection Regulation and its United Kingdom and Swiss equivalents.
| Purpose | Legal basis |
|---|---|
| Providing the Platform, routing and executing requests, and returning Outputs | Performance of a contract |
| Metering, billing, collections, and tax compliance | Performance of a contract and legal obligation |
| Customer and technical support | Performance of a contract and legitimate interests |
| Security, rate limiting, and prevention of abuse and fraud | Legitimate interests |
| Verifying Network Operator hardware and workload integrity | Legitimate interests |
| Sanctions and export control screening | Legal obligation and legitimate interests |
| Reliability engineering and capacity planning, using metadata only | Legitimate interests |
| Website analytics and advertising measurement | Consent where the law requires it, and otherwise legitimate interests |
| Service notices and, where permitted, marketing messages | Legitimate interests, or consent where the law requires it |
| Enforcing our Terms, and establishing or defending legal claims | Legitimate interests and legal obligation |
We do not use Inputs or Outputs to build advertising profiles. We do not make decisions based solely on automated processing that produce legal or similarly significant effects on you, except for the fraud, abuse, and sanctions controls described in Section 12, and you may ask for human review of any of those decisions.
6. How Chutes handles Inputs and Outputs
6.1 API requests
For API requests, our servers may read request content transiently in memory to route and execute a request (for example, to read the model name or to fetch an image for a vision model). We do not write that content to any database, log, or persistent storage, subject to Section 6.3. In end-to-end encryption mode, our servers cannot read the content.
6.2 Chutes Chat, Chutes Search, and other first party applications
Our first party applications store content when storage is part of the product. For example, Chutes Chat stores your conversation history so you can return to it, and Chutes Search may store your query history. We store this content only to provide the application, enforce plan limits, provide support, and prevent abuse, and we never use it to train models. You can delete stored content in the application. Requests from these applications may be served by Confidential Compute Models or by Standard Models. Where an application has a model selector, the selector shows which Models run in a TEE.
6.3 Limited exceptions
Chutes may retain request content on its own systems only in the following circumstances:
- Transient buffers. Content held in memory, a queue, or a streaming buffer for the time needed to complete a request, and never longer than 24 hours.
- Features you enable. Content you save in a first party application, as Section 6.2 describes.
- Abuse and safety investigations. Specific request records reasonably necessary to investigate, report, or stop a credible violation of our Terms or the law, such as child sexual abuse material, credible threats of violence, or attacks on the Platform.
- Legal obligations. Content subject to a legal hold, a valid preservation demand, or a mandatory reporting obligation.
- Your request. Content you send us and ask us to use to reproduce or debug a specific problem.
None of these exceptions applies to requests in end-to-end encryption mode, because in that mode we never hold your content in readable form.
6.4 User Chutes
When you invoke a User Chute that another user created, you run code that its author controls. The author, and not Chutes, decides how that code handles your data, and you should review the author's terms before you send personal data. If you deploy a User Chute that processes personal data, you are responsible for your own privacy notice and legal basis.
6.5 Voice, image, and biometric content
Some Models can generate or transform speech, images, and video. Chutes does not use voice or facial content to identify individuals, does not create or store biometric templates, and does not use biometric content to train models. You are responsible for obtaining every consent the law requires before you submit any person's voice, face, or likeness, including any written release that laws such as the Illinois Biometric Information Privacy Act require. We recommend that you process this kind of content only with a Confidential Compute Model.
7. How long we keep personal data
| Category | Retention period |
|---|---|
| API Inputs and Outputs on Chutes systems | Not retained, subject to Section 6.3 |
| Inputs and Outputs held by Model Providers | As Section 4.1 describes |
| First party application content | Until you delete it, and then removed from backups within 30 days |
| Prompt prefix hashes and routing caches | Generally no longer than 24 hours |
| Usage and billing metadata | 24 months for operations, and 7 years for accounting, tax, and audit records |
| Security, abuse, and access logs, including IP addresses | 90 days, unless needed for an open investigation |
| Sanctions screening records | 5 years from the relevant transaction |
| Account data | For the life of the account and 90 days after it closes, subject to billing and legal retention |
| Support correspondence | 24 months |
| Cookie and analytics identifiers | For the period of each cookie, and no longer than 24 months from your last visit |
| Cryptocurrency transactions on a public blockchain | Permanent and outside our control, as Section 11 explains |
If we are subject to a legal hold, litigation, a regulatory inquiry, or a mandatory reporting obligation, we keep the affected data for as long as that obligation requires.
8. Cookies, analytics, and advertising
8.1 Necessary cookies. We use cookies and similar technologies that are necessary to sign you in, keep the Platform secure, remember your preferences, and enforce rate limits.
8.2 Analytics and advertising. We use Google Tag Manager and Google Analytics to understand how visitors use chutes.ai, and we may use Google advertising tags to measure the effectiveness of our advertising. These technologies run in your browser on our website. They receive information about your visits to the website, and they never receive your Inputs, Outputs, or API usage.
8.3 Your choices. If you visit from the European Economic Area, the United Kingdom, or Switzerland, analytics and advertising storage is turned off until you accept it. Elsewhere, these technologies are on by default, and you can turn them off at any time through the cookie settings on our website or your profile settings. You can also block cookies in your browser.
8.4 Opting out of sharing. Allowing an advertising network to use website data for advertising based on your activity across other websites can be "sharing" for "cross-context behavioral advertising" under the California Consumer Privacy Act and "targeted advertising" under other state laws. You may opt out by turning off advertising cookies in our cookie settings, by using the "Do Not Sell or Share My Personal Information" link in our website footer, by sending a Global Privacy Control signal from your browser, which we treat as a valid opt-out request, or by writing to legal@chutes.ai.
9. When we act for business customers
When a business customer uses the Platform to process personal data about its own users or customers, that business is the controller (or "business" under United States state privacy laws), and Chutes processes that personal data as its processor or service provider under our Data Processing Agreement. If you are an end user of a business that builds on Chutes, please direct your privacy requests to that business. We will help that business respond to your request as our agreement with it requires.
10. How we share personal data
We do not sell personal data. We share personal data only with the following categories of recipients, and only as needed for the purposes in Section 5:
- Model Providers, which receive Inputs and return Outputs when they serve a Standard Model, as Section 4 describes;
- Network Operators, which supply compute for Confidential Compute Models and process your content inside a TEE that is designed to keep it from them;
- infrastructure and service providers, including cloud hosting, content delivery, network security, email delivery, error monitoring, and customer support providers, each under a written agreement that limits its use of the data;
- payment processors and screening vendors, including Stripe, blockchain payment rails, and sanctions and fraud prevention providers;
- analytics and advertising providers, including Google, which receive website data as Section 8 describes and never receive Inputs, Outputs, or API usage;
- professional advisors and affiliates, including lawyers, auditors, accountants, and insurers, under duties of confidentiality;
- parties to a corporate transaction, such as an acquirer or successor in a merger, financing, reorganization, or sale of assets, subject to this Privacy Policy or an equivalent policy;
- authorities, where valid legal process compels disclosure, or where disclosure is reasonably necessary to comply with law, enforce our Terms, prevent fraud or abuse, or protect the rights, property, or safety of any person; and
- anyone else you direct or authorize.
Government requests. We require valid legal process before we disclose customer data, we read requests narrowly, and we object to overbroad demands. Where the law allows, we notify the affected customer before we disclose its data. Because Chutes does not store API content, we generally have no content to produce, and in end-to-end encryption mode we could produce only ciphertext that we cannot decrypt. We cannot respond on behalf of a Model Provider or Network Operator.
11. Payments and blockchain data
Payments are processed by Stripe or through supported blockchain payment rails. We keep only the payment data needed to reconcile balances, issue invoices, and prevent fraud. Blockchain transactions, including wallet addresses, amounts, and timestamps, are public, are copied across independent nodes worldwide, and cannot be changed or erased by Chutes or anyone else. A request to delete or correct personal data does not reach data recorded on a public blockchain, although it does reach the link between your wallet address and your account in our own systems. If you want deletable payment records, use a conventional payment method. We never record your identity, account identifiers, or request content on a public blockchain.
12. Sanctions and export screening
Chutes is a United States company and must comply with United States sanctions and export control laws. To do so, we use your IP address, approximate location, account and organization data, payment data, and, where needed, identity verification data to apply the restrictions in Section 14 of our Terms of Service. We screen account holders and payment addresses against restricted party lists maintained by the United States and other governments, and we may block access from comprehensively sanctioned locations or from proxies used to hide a location. Screening is partly automated. Any denial or suspension is subject to human review on request, which you may ask for by writing to legal@chutes.ai.
13. International data transfers
Chutes is based in the United States and processes personal data there. Our service providers, Network Operators, and Model Providers may process personal data in other countries, including countries in Asia. When we transfer personal data from the European Economic Area, the United Kingdom, or Switzerland to a country that does not have an adequacy decision, we use a lawful transfer mechanism, such as the Standard Contractual Clauses approved by the European Commission, the United Kingdom International Data Transfer Addendum, or their Swiss equivalent, or we rely on another basis that applicable law allows. You may ask for more information about these safeguards by writing to legal@chutes.ai.
14. Your privacy rights
14.1 Your rights. Depending on where you live, you may have the right to:
- know whether we process your personal data and obtain a copy of it;
- correct inaccurate personal data;
- delete your personal data;
- receive your personal data in a portable format;
- object to, or ask us to restrict, certain processing;
- withdraw consent at any time, without affecting earlier processing;
- opt out of the sale or sharing of personal data, targeted advertising, and certain profiling; and
- limit the use of sensitive personal information.
We honor these requests for all users where we are able to, wherever they live, subject to the exceptions the law allows.
14.2 How to make a request. Send your request to legal@chutes.ai from the email address on your account, and tell us which right you want to exercise. We will verify your request, usually by confirming control of your account or email address, and we use verification data only for that purpose. An authorized agent may make a request for you with your signed permission. We respond within the period the law requires, which is one month under the General Data Protection Regulation and 45 days under the California Consumer Privacy Act, and we may extend that period where the law allows.
14.3 What we can reach. We can act on the personal data that Chutes holds, including account, billing, usage, application, and support data. Chutes does not hold API content, so there is ordinarily none for us to provide or delete. We cannot act directly on data held by a Model Provider or Network Operator, but we will forward a valid request concerning Content routed through Chutes to the relevant Model Provider and help as our agreements allow.
14.4 Appeals and complaints. If we decline your request, we will explain why. You may appeal by writing to legal@chutes.ai with "Privacy Appeal" in the subject line, and a person who was not involved in the original decision will review it. You may also complain to your local data protection authority or, in the United States, your state attorney general.
14.5 No discrimination. We will not deny you service, charge you a different price, or give you a different quality of service because you exercised a privacy right.
15. Information for California residents
In the past 12 months, we collected the following categories of personal information described in the California Consumer Privacy Act: identifiers (such as name, email address, IP address, and account identifiers); customer records and commercial information (such as billing records and purchases); internet or network activity (such as usage metadata and website interactions); approximate geolocation derived from IP addresses; the content of your Inputs, which may contain any category of information you choose to include; and inferences only to the extent needed for fraud prevention and security. We collect this information from the sources in Section 3, for the purposes in Section 5, and keep it for the periods in Section 7. We disclose it for business purposes to the categories of recipients in Section 10. We do not sell personal information. We share website data for cross-context behavioral advertising only as Section 8 describes, and you may opt out at any time. We do not knowingly sell or share the personal information of anyone under 16. We do not use or disclose sensitive personal information for purposes other than those the law permits without a right to limit.
16. Security
We use technical and organizational measures designed to protect personal data, including encryption in transit, encryption at rest, hardware memory isolation for Confidential Compute Models, hardware attestation, verification of workload images, restriction of network access for running workloads, role based access controls, multifactor authentication for administrative access, and an incident response process. No system is completely secure, and we cannot promise that the Platform will never be compromised. You are responsible for protecting your credentials and API keys. If a personal data breach occurs, we will notify affected individuals and regulators as the law requires. Please report suspected vulnerabilities to support@chutes.ai with "Security Report" in the subject line.
17. Artificial intelligence transparency
Outputs come from statistical models and may be inaccurate, incomplete, or misleading. You are responsible for any labeling or disclosure that the law requires when you publish content generated with the Platform, including under Article 50 of the European Union Artificial Intelligence Act where it applies.
18. Children
The Platform is intended for adults. You must be at least 18 years old to use it. We do not knowingly collect personal data from children. If you believe a child has given us personal data, please contact legal@chutes.ai, and we will delete it.
19. Changes to this Privacy Policy
We may update this Privacy Policy from time to time. We will post the updated version at chutes.ai/privacy and change the "Last updated" date. If we make a material change, we will notify you by email or through the Platform before the change takes effect.
20. Contact us
Questions or requests about this Privacy Policy may be sent to legal@chutes.ai or to Chutes Global Corp., 2810 N Church St., Suite 89723, Wilmington, Delaware 19802, United States.
