Privacy policy

Who we are

This policy explains how Inference Bureau ("we", "us") handles personal data when you visit the website or use the console, the API, Auditor and the rest of the service. For data inside your requests, your team decides what to send and why; we process it on your team's behalf.

What we collect

Account details. Your name, email address and, if you add one, a profile picture; your password (stored only as a hash), passkeys and two-factor settings; and the teams you belong to and your role in each.

Sign-in and security records. For each session, the IP address, browser and device it came from and when it was active, so you can see and end sessions and so we can spot abuse.

API keys. We store a hash of each key with its prefix and last four characters, never the full key.

Usage records. For each request: the time, team, project, key, model, provider, token counts, latency, cost, any tags, and the end-user ID if your application sends one. These records are what Logs, spend and billing are built from.

Request content. The prompts, files and outputs in your requests pass through the service to reach a model provider. They are not stored unless an owner or admin turns on the Log content Directive. Then each request's body and response is encrypted with a key specific to your team, viewable only by your team's owners and admins, and deleted after 7 or 30 days, whichever your team chose, or as soon as Log content is turned off. Responses you ask the API to store are kept until they expire or you delete them.

Billing. Your team's top-ups, balance, statements and refunds. Card details are collected and held by Stripe; we keep only the card's brand, last four digits and a fingerprint Stripe provides, which we use to give starter credit once per card.

Integrations. When you connect GitHub, Shopify, Linear or Jira, we receive what you authorize: for Auditor, the repositories, pull requests and code it reviews, theme files, and the issues it files. Secrets you give an integration are stored encrypted.

Suggestions and feedback. What you send through the Roadmap or a Beta badge, with the page you were on. Pictures attached to them are stored at an address anyone who has it can open.

Emails to us. Whatever you include when you write to us.

How we use it

• To run the service: authenticate you, route requests, apply your Directives and budgets, and show you Logs and spend.
• To bill your team and handle payments, credits and refunds.
• To keep the service secure: detect abuse, investigate problems and enforce our Terms of service.
• To send account emails: verification, password resets, invitations, receipts, low-balance alerts and notices about changes to the service or these policies.
• To answer you when you contact us, and to follow up on suggestions.
• To improve the service using usage records in aggregate. We don't use your request content for this, and we don't use it to train models.

We don't sell personal data, and we don't use it for advertising.

Model providers

To answer a request, we send its content to the model provider it's routed to. The provider handles it under its own terms and privacy policy; the Providers page lists who they are. Your team's Directives, such as Require ZDR and Require no training, restrict routing to providers that have made those commitments.

Linear

When a team connects Linear to Auditor, an admin of the Linear workspace authorizes the Auditor app. It asks for permission to read and write issues and comments, and to be mentioned in and assigned issues. What it does with that:

• What it reads. The issue a pull request names, or the issue you mention or assign Auditor on: its identifier, title, description, URL and the links attached to it, the text of your request, the name of who asked and, when your team limits who may have Auditor change code, their email address (to check they're allowed). Auditor doesn't read other issues, projects or members of the workspace.
• What it writes. A comment on the issue when a pull request misses something the issue asks for, edited on later reviews, and its replies and progress in the Linear agent session you start by mentioning or assigning it. When an automation is set up to, it opens an issue in the Linear team you name. Code changes it makes go to your GitHub repository, not to Linear.
• How it's used. Only to do what you asked: check the pull request against the issue, and answer or work on the issue. The issue's text is sent to the model provider Auditor's request is routed to, as described under Model providers. It isn't used to train models or sold, and it goes to no one beyond that provider and the service providers below.
• How it's stored. The access token is stored encrypted. An issue's identifier, title, the start of its description, the request and who made it are kept with the Auditor run they started, which your team sees in Auditor's activity log, for as long as your team exists or until you ask us to delete them.
• Disconnecting. Disconnect Linear on its page under Add-ons › Issue trackers in Auditor to delete the stored token, and remove the Auditor app in Linear's workspace settings to revoke its access. Email us to have the runs it started deleted too.

Service providers

We use these companies to run the service, and they process personal data only on our instructions:

• Vercel: hosting, scheduled jobs, file storage, isolated build environments and delivering requests to model providers
• Cloudflare: DNS, custom domains and request routing for apps and theme sites hosted on Inference Bureau
• Neon: the database
• Upstash: caching and background job queues
• Stripe: payments
• Resend: email delivery
• GitHub, Shopify, Linear and Jira: only when you connect them

We may also disclose data when the law requires it, to protect the rights and safety of our users or the public, or as part of a merger, acquisition or sale of the service, in which case this policy continues to apply to it.

Cookies

We use cookies to keep you signed in and to secure your session. They expire after 7 days without use. We don't use advertising or analytics cookies, and no third-party trackers run on the site.

How long we keep it

• Account details for as long as your account exists.
• Request content, when Log content is on, for the 7 or 30 days your team chose.
• Usage records for as long as your team exists, because spend history and statements are built from them.
• Billing records for as long as tax and accounting law requires, even after an account is deleted.
• Sign-in records until the session ends or expires.

When you delete your account or a team, we delete its data or make it anonymous, except what we must keep by law or to resolve disputes.

Your choices and rights

• Correct your name, email and profile picture in your account settings.
• Delete your account from your account settings. If you own a team, transfer it or delete it first.
• Export your Logs and spend as CSV from the console.
• Ask for a copy of your data, a correction, deletion, or to object to or restrict how we use it, by emailing us. We'll answer within 30 days, and may need to confirm it's you first.

Depending on where you live, you may have further rights under laws such as the GDPR or the CCPA, including the right to complain to your data protection authority. We honor them for everyone, wherever you are.

If you are an end user of an application built on Inference Bureau, contact that application's developer first: they decide what's sent to us.

Security

Passwords and API keys are stored only as hashes, request content and integration secrets are encrypted, and connections use TLS. Access inside the team is limited to people who need it to run the service. No system is perfectly secure; if a breach affects your data, we'll tell you as the law requires.

Children

The service is not for anyone under 18, and we don't knowingly collect data about children.

International transfers

We and our providers process data in the United States and other countries. Where the law requires it, transfers are covered by safeguards such as the European Commission's standard contractual clauses.

Changes to this policy

When we change this policy, we'll update the date at the top. If a change is material, we'll tell you by email or in the console before it takes effect.

Contact

Questions or requests about your data: hello@inferencebureau.com.