Data Processing Agreement
Last updated November 17, 2026
This Data Processing Agreement ("DPA") is part of the Terms of Service between Zalmir LLC ("Zalmir", "we") and a customer that uses Zalmir for its business ("you"). It applies automatically, without needing to be signed, whenever we process Customer Personal Data for you, from when you accept the Terms. Where this DPA and the Terms differ about Customer Personal Data, this DPA applies.
What it covers
- Customer Personal Data is personal data in the content you send to Zalmir through the API or Chat, and in the responses: prompts, inputs, files and outputs, where data protection laws apply to it.
- For Customer Personal Data, you are the controller (or a processor acting for your own customer), and Zalmir is your processor and your "service provider" under the California Consumer Privacy Act.
- Your account, sign-in, billing and usage records, security logs and support messages are not Customer Personal Data. Zalmir handles those as a controller, under the Privacy Policy.
- Data protection laws means the laws that apply to the processing, including the EU General Data Protection Regulation ("GDPR"), the UK GDPR and Data Protection Act 2018, the Swiss Federal Act on Data Protection, and the California Consumer Privacy Act and other US state privacy laws.
How we process it
- We process Customer Personal Data only to run your requests and provide the service, and only on your instructions. Your instructions are the Terms, this DPA, how you use the service (including the models you choose and your account settings) and any other reasonable instructions you give us in writing that fit the Terms. If we think an instruction breaks data protection law, we will tell you.
- API requests and answers are not stored after the request is complete, except what is held briefly in memory or in transit to complete and secure it, and anything the law requires us to keep. Chat defaults to Saved to your account: on your instruction, we store readable conversation titles, prompts, responses, model reasoning, model choices, timestamps and request IDs until you delete them or close your account, so you can continue on any device. Our database provider encrypts this data at rest. You can choose This device only, where history stays in your browser and Zalmir never stores it. Saved content is accessible only to run and secure the service, never sold, never used for training, and excluded from logs, analytics and session recordings.
- We do not use Customer Personal Data to train models. We do not sell or share it (as the California Consumer Privacy Act defines those words), combine it with other data, or use it for any purpose other than providing the service to you.
- You are responsible for having a lawful basis, and giving any notices or getting any consents the law requires, for the Customer Personal Data you send. Send special categories of data, or other sensitive data, only where you need to and the law allows it.
- Schedule 1 describes the processing.
Model providers and independent providers
- Each request runs on the model you choose, served by a model provider or, where we offer them, an independent provider. The model provider or independent provider for each model is shown on Zalmir, by name or, for an independent provider we list without its name, as "Independent". Sending your request to it is part of your instruction. If your saved routing policy allows another provider when the first one fails, that provider may run the request instead.
- Each model provider or independent provider handles the content it receives under its own terms and data policy, which Zalmir does not control (see Model providers and independent providers in the Terms). Choose models with this in mind.
- To the extent data protection law treats a model provider or independent provider as our subprocessor, you authorize the model providers and independent providers that run the models you choose. Our responsibility for them is as set out in the Terms, to the extent the law allows.
Subprocessors
You authorize these services, which run parts of Zalmir for us, to process Customer Personal Data on our behalf:
- Vercel Inc.: website and API hosting. Requests are handled in Frankfurt, Germany, and pass through Vercel's worldwide network.
- Supabase Inc.: database and API functions, including the gateway that forwards each request to its model provider or independent provider. Frankfurt, Germany.
- Google LLC (Google Workspace): our email, where you send us Customer Personal Data by email. United States and other countries.
- Resend: emails sent by our website, such as contact form messages. United States.
We use each under written terms that include data protection commitments covering the data it processes for us. We will update this list here with a new date before a new subprocessor processes Customer Personal Data. If you want to be told by email at least 30 days in advance, ask at support@zalmir.ai. You may object to a new subprocessor on reasonable data protection grounds within 30 days of that notice; we will then try to find a reasonable solution, and if we cannot, you may stop using the affected part of the service.
Our people
Only people who need access to provide the service can reach Customer Personal Data, and they are bound to keep it confidential.
Security
We protect Customer Personal Data with technical and organizational measures suited to the risk, including those in Schedule 2, and we update them as the service changes.
Security incidents
If a breach of security leads to the accidental or unlawful destruction, loss, alteration, unauthorized disclosure of, or access to Customer Personal Data processed by us or our subprocessors, we will tell you without undue delay, and within 72 hours of becoming aware of it. We will give you the information you need to meet your own obligations as it becomes available, and take reasonable steps to contain the breach and limit its effects. If we learn of such a breach at a model provider or independent provider that ran your requests, we will tell you what we know. Telling you about a breach is not an admission of fault.
Helping you
- If someone asks us to exercise their data protection rights about Customer Personal Data, we will tell you, unless the law forbids it, and will not answer them ourselves except to direct them to you. API content is not retained, but saved Chat history may contain their data. We will give you reasonable help to respond and to export or delete saved Chat content.
- We will give you reasonable help, with information we hold, with data protection impact assessments and consultations with data protection authorities about the service.
Government requests
If a public authority asks us for Customer Personal Data, we will tell you, unless the law forbids it, and disclose only what we are legally required to. We will challenge a request we reasonably believe is unlawful, where that is reasonable and possible.
Deletion
API request content is not kept after each request. You can export saved Chat history as JSON and delete individual conversations or all saved chats from Chat; deletion removes them from the active database. Closing your account also removes saved Chat history. Only content-free internal deletion markers remain to prevent late requests from restoring deleted chats; these markers are removed when your account closes. When the Terms end, we delete any Customer Personal Data we still hold within 30 days, unless the law requires us to keep it.
Audits
We will answer your reasonable written questions about how we meet this DPA. If written answers are not enough and data protection law gives you the right, you may have Zalmir's own systems audited once a year by an independent auditor bound by confidentiality, at your own cost, with 30 days' notice and during business hours. More frequent audits are allowed where a data protection authority requires them or after a security incident.
International transfers
Zalmir LLC is in the United States. Customer Personal Data is processed mainly in Germany, and may also be accessed from or transferred to the United States and other countries where we or our subprocessors work. Where data protection laws require safeguards for these transfers:
- From the European Economic Area: the Standard Contractual Clauses approved by the European Commission in Decision 2021/914 apply and are made part of this DPA: Module 2 (controller to processor) where you are a controller, and Module 3 (processor to processor) where you are a processor. Clause 7 (docking) applies; under Clause 9(a), Option 2 (general authorization) applies, with 30 days' notice; the optional wording in Clause 11 does not apply; under Clauses 17 and 18, the law of Ireland applies and the courts of Ireland decide disputes. Annex I is Schedule 1, Annex II is Schedule 2, and Annex III is the subprocessor list.
- From the United Kingdom: the International Data Transfer Addendum to the Standard Contractual Clauses issued by the UK Information Commissioner (in force from March 21, 2022) applies, completed with the details in this DPA. Either of us may end it as Section 19 of the Addendum allows.
- From Switzerland: the Standard Contractual Clauses apply as above, with the Swiss Federal Data Protection and Information Commissioner as the supervisory authority, references to the GDPR read as references to Swiss law, and people in Switzerland able to bring claims there.
If the Standard Contractual Clauses and this DPA or the Terms differ, the Standard Contractual Clauses apply.
Liability
Each party's liability under this DPA is subject to the Limitation of liability in the Terms, to the extent the law allows. Nothing in this DPA limits the rights of the people the data is about under the Standard Contractual Clauses.
Changes
We may change this DPA at any time. The new version applies from when we post it here with a new date. If a change materially affects you, we will also tell you by email or in the Console. We will not change the Standard Contractual Clauses or the UK Addendum except as they allow.
Contact
Email support@zalmir.ai about this DPA, Customer Personal Data or a security concern.
Schedule 1: Details of processing
- Data exporter: the customer, as identified by its Zalmir account. Contact: the email address on the account. Role: controller, or processor for its own customer.
- Data importer: Zalmir LLC, a Delaware limited liability company, United States. Contact: support@zalmir.ai. Role: processor.
- Activities: providing the Zalmir API service under the Terms. Both parties accept this DPA, and these details, when the customer accepts the Terms.
- People the data is about: whoever the customer's content is about, such as its users, customers and staff, and other people named in the content.
- Kinds of personal data: whatever the customer includes in prompts, inputs, files and outputs, and any identifiers it puts in request fields, such as an end-user ID.
- Sensitive data: none is intended. If the customer sends any, the measures in Schedule 2 apply, including the choice of saved Chat or device-only history and the API no-storage policy.
- How often: continuously, for each request the customer sends.
- Nature of the processing: receiving each request, counting its tokens to charge for it, forwarding it to the model provider or independent provider that serves the chosen model, returning the response, and retaining Chat history when you choose saved storage.
- Purpose: providing the service under the Terms.
- How long: API content is not kept after the request is complete; saved Chat history is kept until you delete it or close your account (see How we process it); processing continues for as long as the Terms apply.
- Subprocessors: those in the subprocessor list, for the same processing and for as long as the Terms apply.
- Supervisory authority: the authority of the EU country where the customer is established; if it is not established in the EU but has appointed an EU representative under the GDPR, the authority where that representative is; otherwise, Ireland's Data Protection Commission.
Schedule 2: Security measures
- Encryption: the website and API accept only encrypted connections (HTTPS), and requests travel to model providers and independent providers over encrypted connections. Our database host encrypts stored data.
- Content storage: API requests and answers are not written to our database or logs. Saved Chat history is stored in owner-isolated tables, with server-only access limited to running and securing the service, until you delete it. This device only history is not stored on our servers. Chat content and titles never appear in logs, analytics or session recordings. Error records for API requests hold only the kind of error and the route.
- API keys: we store only a one-way hash of each key and a short fingerprint, never the key itself.
- Access control: our database is closed by default. Each table is protected by row-level security and can be reached only by our own server functions, with the least access each needs. Payment records can only be added to, never edited or deleted, and the public database roles cannot empty tables.
- Secrets: service keys are kept in our hosting provider's encrypted settings, never in our code, and each release of the website is checked for leaked secrets before it is published.
- Sign-in: handled by our sign-in service. Staff access is limited to verified company accounts.
- Abuse protection: API keys are checked on every request, and the API has rate limits.
- Location: the API and the database run in Frankfurt, Germany.
- Testing: changes to payments, access rules and the API are covered by automated tests.
- Incidents: we investigate suspected breaches and tell affected customers as set out in Security incidents.