Privacy Policy

1. Overview

Data Workers, Inc. ("we", "us", "our") operates the dataworkers.io website, publishes the Data Workers open-source software, and offers paid and Enterprise services (collectively, the "Services"). This Privacy Policy explains what information we collect, how we use it, and your rights regarding that information.

This policy describes what our software and our website actually do. Where a control is not implemented, we say so rather than describing the version we would prefer.

2. How the Services Are Deployed, and Why It Matters

Most of what follows depends on where the software runs, so it is worth stating up front:

• •The dataworkers.io website is operated by us. Everything in Sections 4 and 11 applies to it.
• •The open-source software, when you self-host it, runs on your own infrastructure, under your own credentials, inside your own network. In that mode we do not receive your queries, your results, your schemas, or your credentials, because the software does not send them to us. What it does send to third parties is described in Section 6, and that is a decision you control through configuration.
• •Managed, hosted, and Enterprise Design Partner arrangements are governed by the written agreement covering them, which takes precedence over this policy where the two differ.

Sections 5, 6, 7 and 10 describe the behavior of the software itself. They apply wherever it runs, including on your own infrastructure, because you need them to assess the software before you deploy it.

3. Customer Data Content vs Operational Metadata

This is the distinction that matters most, so we state it plainly rather than burying it.

The software does store the content of results, not only metadata about them. Anyone evaluating it should plan for that. Specifically, when the relevant features are used, the software writes the following to the storage you have configured:

• •Query results are cached for 24 hours, keyed to the question that produced them, including the returned rows.
• •Dataset profiling stores sample values. When a dataset is profiled for the catalog, up to five of the most frequent actual values per column, plus minimum and maximum values, are written into the catalog graph. These are real values from your data and they are not redacted before storage.
• •Analysis view caching materializes full result sets to local disk for reuse. These cache files are not currently expired on a schedule.
• •Schema details - table and column names, types, and row counts - are cached and stored in the catalog.
• •Audit and enrichment records may include a truncated extract of tool output alongside the tool name, parameters, outcome, and timing.

Operational metadata - which tool ran, its parameters, success or failure, duration, and timestamps - is recorded separately from the above.

Because this storage is on infrastructure you configure and control when you self-host, retention and deletion of that content are governed by your configuration and your own retention policy, not by ours. If you need result content not to be retained, review the caching and profiling behavior before you enable those features.

We do not use customer data content to train models, and we do not use it to improve the product beyond serving the immediate request that produced it. We have no model-training pipeline that ingests customer data. Section 6 describes what is sent to third-party model providers to serve a request, and what those providers may do with it.

4. Information the Website Collects

4a. Analytics

We use Google Analytics 4 on dataworkers.io. It loads on every page. It records pages visited, referral source, device type, browser and operating system, approximate country and city level location, and a small number of events we define: page views, scroll depth on articles, and submission of the demo request form. Google Analytics 4 does not log or store IP addresses in the way earlier versions did, but IP address is processed transiently by Google to derive approximate location. Google's own privacy terms govern what Google does with this data.

We also use Amplitude for product analytics. It records the pages you view, the links and buttons you click (including their visible labels), and when a form is started or submitted. It does not record what you type into form fields, and we do not record session replays. Amplitude assigns your browser a random device identifier; we never link it to your name or email address. Amplitude's own privacy terms govern what Amplitude does with this data.

If your browser sends a Global Privacy Control signal, neither Google Analytics nor Amplitude loads.

Google Analytics and Amplitude load only after you choose Accept in the cookie banner. Until you choose, and if you choose Reject, neither loads and neither sets cookies. You can change your choice at any time with the Cookie settings link at the bottom of every page.

Two measurements run without cookies and without storing anything in your browser, so they do not depend on the banner. Cloudflare Web Analytics counts page views and page load performance; it does not use cookies, local storage or fingerprinting, and Cloudflare's own privacy terms govern it. Our own page experiments show different versions of an offer on some article pages: our server picks the version from a one-way hash of your IP address, browser user agent and a value that changes every day, and logs which version was shown and whether its link was clicked. The hash cannot be reversed into your IP address, the IP address itself is not stored, and the logs are kept for no more than 30 days.

4b. Forms

If you submit a demo request, contact form, Design Partner application, or the deletion request form below, we collect what you type into it: typically your name, email address, company, and any message or reason you provide. Submissions are delivered to our internal email inbox through our email delivery provider. We use them to respond to you.

4c. Abuse prevention

Our form endpoint applies a per-IP submission throttle. To do this it stores your IP address for ten minutes and then discards it. If a submission trips an abuse guard, we log the IP address, network operator, and browser user agent so that we can identify and block abusive traffic. We do not use this information for any other purpose, and we do not use it to build a profile of you.

5. Credentials and Secrets

The software does not manage secrets for your external systems, and we do not receive your warehouse credentials. Connectors read credentials from the environment of the machine the software runs on. Data Workers does not store or transmit them to us.

This means you are responsible for securing those credentials using your own secrets management, and for the scopes and permissions you grant. Our setup tool, if you use it, writes the credentials you supply to a plaintext environment file on your own machine, which you should protect accordingly. Grant the narrowest permissions that let the agents do the job you want done.

On logging: we found no code path that writes credential values to logs. Logs record whether a credential is present, not what it is. One internal development harness logs the first eight characters of a webhook signing secret; that harness is not part of a normal deployment. We describe this precisely rather than claiming a blanket guarantee.

On encryption: because warehouse credentials are never persisted by the software, there is no credential store for us to encrypt, and we make no encryption-at-rest claim about them. The software does not itself enforce TLS or other network-level controls; transport security is a property of your deployment and of the connections your infrastructure makes. Where you deploy it, terminate TLS and apply network controls using standard practice.

6. AI Model Providers, Including OpenAI

Data Workers agents call large language models. This is the point at which your data can leave your own infrastructure, so read this section carefully.

Which provider is used is determined by which API key is configured in the environment where the software runs. Anthropic is selected first when an Anthropic key is present; OpenAI is used when only an OpenAI key is present. If no model provider key is configured, no external model call is made at all. That choice is yours, not ours, when you self-host.

Content of your data, not only schema, is sent to the model provider. Depending on which features you use, this can include: up to fifty rows of query results, serialized, in order to generate an explanation of those results; up to five actual values from a column being classified for personal information; extracts of record text; your table and column names, row counts, the generated SQL, and the question you asked. There is no redaction or masking layer between your data and the model request. If a column contains personal information and a feature sends sample values from it, that personal information is in the request.

The software includes a mode that blocks external model calls entirely and keeps processing local. It is off by default and must be turned on deliberately.

6a. What OpenAI does with data, and what we control

Once data reaches a model provider, that provider's terms govern it, not ours. The distinction that matters:

• •Consumer ChatGPT - content may be used to train OpenAI's models unless you opt out in your ChatGPT settings.
• •ChatGPT Business, Enterprise, Edu, and the OpenAI API - business data is not used to train OpenAI's models by default.

If you connect Data Workers to an AI client such as an assistant or coding tool, that client sends tool results to its own model provider, and which of the two categories above applies depends on the plan you are on with that provider. We do not control it and cannot change it from our side. Review OpenAI's privacy policy, OpenAI's terms of use, and Anthropic's privacy policy for the provider you use.

7. How We Use Information

• •To respond to demo requests, contact messages, and Design Partner inquiries
• •To understand how the website is used and improve its content
• •To operate the Services, including executing the tools you invoke and returning their results
• •To detect, investigate, and prevent abuse, unauthorized access, and unauthorized automated access
• •To comply with legal obligations

We do not sell your personal information. We do not use your data for advertising. We do not use customer data content to train models.

8. Third Parties and Subprocessors

The following third parties process data in connection with the website and our own operations. We list only those we can evidence:

• •Cloudflare - hosting, CDN, and the form endpoint for dataworkers.io
• •Google (Analytics) - website analytics
• •Amplitude - website product analytics
• •Resend - delivery of form submissions to our inbox
• •Google (Gmail) - the inbox that receives form submissions
• •GitHub - hosting of the open-source code repository
• •Anthropic and OpenAI - model providers, where a key for them is configured in a deployment, as described in Section 6

Where you self-host, you choose your own infrastructure and model providers, and they are your subprocessors rather than ours.

9. Data Retention

• •IP addresses used for form throttling - ten minutes.
• •Form submissions and the correspondence they start - retained in our inbox for as long as needed to respond and to keep a record of the relationship, and deleted on request.
• •Analytics data - retained according to the retention settings on our Google Analytics property and our Amplitude project.
• •Query result cache - 24 hours.
• •Protocol session records - 30 minutes of inactivity. These records contain a session identifier and timestamps only.
• •Catalog profiles, analysis view caches, and audit records - retained until you delete them. These are not currently expired on a schedule, so their lifetime is set by your own retention policy and housekeeping.

We state the last item explicitly because it is the one most likely to surprise you. If indefinite retention of profiled sample values or cached result sets is not acceptable in your environment, configure or disable those features accordingly.

10. Security

Measures that are in place, described accurately:

• •The dataworkers.io website is served over HTTPS with HTTP Strict Transport Security, a content security policy, clickjacking protection, and MIME type sniffing protection.
• •The form endpoint applies rate limiting, a hidden field trap for automated submissions, and validation before any message is sent.
• •We do not receive your warehouse credentials, which removes a class of risk entirely rather than mitigating it.
• •Optional encryption at rest is available for audit logs and must be enabled with a key you supply. It is not on by default.
• •The software supports an offline mode that blocks external model calls. It is off by default.

Limitations we are not going to leave out. The software is pre-release. Its default in-memory and local implementations are intended for development and evaluation and do not provide durability, encryption at rest, or access controls. The software does not itself enforce TLS or network-level controls. Substantial portions of it were developed with the assistance of generative AI tools and it should be treated as experimental. Configure production infrastructure, transport security, and access controls yourself before relying on it.

We hold no security certification and make no certification claim. We have not published an independent audit, and we do not carry a compliance attestation. Do not infer one from anything on this page.

11. Cookies and Similar Technologies

This site uses cookies and browser storage for analytics, which only run if you accept them, and for a few site features. We do not run advertising pixels or third-party ad trackers on this site.

• •Google Analytics - cookies named _ga and _ga_ followed by our property ID, to distinguish visitors and sessions. They last about 13 months.
• •Amplitude - cookies named AMP_ and AMP_MKTG_ followed by our project ID, plus related browser storage entries, holding a random device identifier and session state. They last about 12 months.
• •How you found us - our own code stores the link or campaign that first and most recently brought you to the site in your browser's local storage (dw_attr_first, dw_attr_last). It is sent to us only if you submit a form, so we know how you heard about us.
• •Your cookie choice - whether you chose Accept or Reject is kept in local storage (dw_consent) for 12 months, so we do not ask again on every page.
• •Preferences - if you mute the site's interface sounds, that choice is remembered in local storage.

Cloudflare Web Analytics and our page experiments set no cookies and write nothing to your browser.

The Google Analytics and Amplitude cookies are set only after you accept. If your browser sends a Global Privacy Control signal, they are never set and the banner does not ask. If you accepted and later choose Reject in Cookie settings, we delete those cookies from your browser. Rejecting analytics does not affect how the site works.

12. Your Rights, and Legal Bases in the EEA and UK

Depending on where you live, you may have the right to access the personal data we hold about you, to have it corrected or deleted, to receive it in a portable format, to object to or restrict processing, and to withdraw consent where we rely on it.

For users in the European Economic Area and the United Kingdom, we process personal data on these bases:

• •Contract - to provide the Services you asked for and to respond to your inquiry.
• •Legitimate interests - to secure the Services, to prevent abuse and unauthorized automated access, and to understand how our website is used.
• •Consent - for analytics cookies, which load only after you accept in the cookie banner. You can withdraw consent at any time through Cookie settings.
• •Legal obligation - where the law requires us to retain or disclose information.

To exercise any of these rights, use the form in Section 17 or contact us using the address in Section 15. We will respond within 30 days. You also have the right to complain to your local data protection authority.

13. California Privacy Rights

This section applies to California residents under the CCPA as amended by the CPRA. In the preceding twelve months we have collected the following categories of personal information:

• •Identifiers - name, email address, company, and IP address. Source: you, through our forms; and your browser, through the website. Purpose: responding to you, and abuse prevention. Retention: as set out in Section 9.
• •Commercial information - the content of an inquiry, including what you tell us about your company and your interest. Source: you. Purpose: responding to you. Retention: as set out in Section 9.
• •Internet and network activity - pages visited, referral source, scroll depth, device, browser, operating system, and approximate location. Source: your browser, through Google Analytics and Amplitude. Purpose: understanding website usage. Retention: per our analytics settings.

We do not sell personal information, and we have not sold personal information in the preceding twelve months. We do not knowingly collect or process sensitive personal information through this website, and we do not use personal information for cross-context behavioral advertising. We honor Global Privacy Control: if your browser sends it, our analytics tools do not load.

You have the right to know what we collect, to delete it, to correct it, to obtain a portable copy, and not to be discriminated against for exercising those rights. Use the form in Section 17 or the address in Section 15.

14. International Transfers, Children, and Separate Agreements

International transfers. We are based in the United States and the third parties listed in Section 8 process data in the United States and other countries. If you access the Services from the European Economic Area or the United Kingdom, your personal information will be transferred outside your home jurisdiction. Where a transfer requires a safeguard, we rely on the mechanisms our providers make available, including standard contractual clauses where they offer them.

Children. The Services are not directed to children, are intended for business use, and we do not knowingly collect personal information from anyone under 16. If you believe a child has provided us personal information, contact us and we will delete it.

Separate agreements. Paid customers and organizations enrolled in the Enterprise Design Partner program may be covered by a separate written agreement, which may include a data processing agreement. Where such an agreement exists and conflicts with this policy, that agreement governs. Nothing in this paragraph should be read as confirming that any particular agreement is already in place; ask us and we will tell you what applies to you.

15. Contact

For privacy questions or to exercise your rights, contact us at hello@dataworkers.io. For legal and partnership matters, contact partners@dataworkers.io, which is the address used in our Terms of Use.

16. Changes to This Policy

We may update this Privacy Policy from time to time. Changes will be posted on this page with an updated date. Continued use of the Services after changes constitutes acceptance of the revised policy.