Trajector

For contributors·v1.1 · Effective 21 August 2026·All legal documents

In short

  • We collect your account details, a hash of the network you signed up from (deleted after 90 days), sessions from the projects you explicitly enabled, the devices you pair, your withdrawal address, upload records, and server logs.
  • Scrubbing targets secrets and credentials, not personal data. File paths, file contents and command output are kept as they were sent, and a path normally carries your user name.
  • Your account information is not part of the data we license to purchasers and is not sold.
  • Withdrawals are stablecoin transfers on public blockchains: the destination address and the amounts are permanently visible to anyone.
  • You can read and delete your sessions from the My data page. Closing an account is handled by hand, and your authorization record is kept either way.
  • Session content is stored in Hong Kong and our account emails are sent from the United States, so your data crosses borders.

This summary is not part of the document and is not what you accept or rely on. Where it and the text below differ, the text below governs.

This policy covers your account information. What you license Trajector to do with the coding sessions themselves is a separate document — the Data Collection and Use Authorization. If you are evaluating Trajector data as a purchaser, the document written for you is the Data Authorization Mechanism Statement, not this page.

Trajector Privacy Policy (v1.1)

Last updated: October 20, 2026. Trajector is operated by PublicAI Foundation, a foundation company incorporated in the Cayman Islands. Questions about this policy, or a request about your data: support@publicai.io.

What we collect

  • Account information. Your email address, and a hash of your password (never the password itself) if you sign up with email. If you sign up with Google or GitHub, we receive the email address and the account identifier those providers share. From GitHub we also receive public profile information, which we use to decide whether an account can be admitted to the beta without a manual review.
  • Sign-up network. When an account is created, we store a keyed hash of the IP address it was created from, and a keyed hash of that address's network (the /24 block for IPv4, the /48 block for IPv6). We do not store the address itself. Our staff use these hashes only to see whether several accounts were created from the same place, as one sign of one person holding more than one account. Each record is deleted 90 days after the account was created.
  • Coding session data. Only from projects you have explicitly enabled for collection. The next section says what a session contains.
  • Devices and payouts. The devices you pair with your account, and the wallet address you provide for withdrawals.
  • Upload records. For each upload: the time, the device, the CLI version, the size, our response, and the user-agent string of the client that sent it.
  • Your authorization record. When you accept the Data Collection and Use Authorization we record what §6 of that document lists, including the name and the jurisdiction you declare, your IP address and your browser's user-agent string. That record cannot be edited afterwards, and it outlives your account for the reason §6 gives.
  • Operational records. Server logs kept for security, debugging and abuse prevention.

What a coding session contains

  • Scrubbing targets secrets, not personal data. The client scrubs secrets and credentials on your machine before anything is uploaded. After upload we scan again for the same kinds of value, plus email addresses and telephone numbers. That second scan holds a session it hits for review by our staff. It does not mask part of one. A held session earns nothing and is not delivered; it is released only if a reviewer confirms the hit was a false positive.
  • Some addresses pass the scan by design. Addresses that belong to automation, such as GitHub users.noreply.github.com addresses, are allowed through. A GitHub noreply address contains a GitHub user name.
  • Everything else in a session is kept as it was sent. That includes absolute file paths and working directories, which normally contain your operating-system user name and your repository names; the contents of files a tool read or wrote; and the output of commands that were run. Tool output is also part of what a session is paid for.
  • Personal data about other people is not in scope. If a session quotes a colleague's name, an author line or a commit author, nothing in the scrubbing removes it. Before we publish anything as an open dataset we review it again for personal data, but that review covers the selected portion only. It is not a statement about everything we hold.

You can read every session we hold on the My data page, in the form we hold it.

How we use it

  • To operate your account: sign-in sessions, email verification, and the closed-beta review of new sign-ups.
  • To review, accept and credit your uploads, and to show them to you on the My data page.
  • To commercialize desensitized session data as described in the Data Collection and Use Authorization, including its use for AI model training.
  • To detect and act on abuse (fabricated sessions, bulk farming, one person holding more than one account).
  • To handle requests from people who say a session contains their work. We do not give a contributor's identity to the person making such a request unless a court or the law requires us to.

Cookies and third-party requests

  • trajector_session keeps you signed in. It is httpOnly, SameSite=Lax, and it lasts seven days.
  • trajector_oauth exists only while a Google or GitHub sign-in is in progress, to protect that redirect. It is httpOnly, SameSite=Lax, and it lasts ten minutes.
  • The site loads Google Analytics. It sets its own cookies and reports page views and a few product events (such as "signed up" or "paired a device") to Google. Those events carry no personal data: no email address, account id, device name or wallet address. We use it to see how the product is used. There are no advertising pixels on the site, and we do not sell or share this data. Google Analytics is the only third-party request the site makes, and it is not loaded on the authorization page: /authorization sends no request to Google at all, whether you open it directly or reach it from another page of the site. Everything else a page loads, including the handwriting fonts on the authorization page, is served from Trajector's own domain.

Sharing, and where your data is held

  • Desensitized session data is licensed to customers as authorized by the Data Collection and Use Authorization. Your account information (such as your email address) is not part of that data and is not sold.
  • A selected portion of the desensitized session data may also be published as an open dataset, under the Open Data Commons Attribution License v1.0 and Creative Commons Attribution 4.0 International. No contributor identity is published with it.
  • If you sign in with Google or GitHub, those providers process the sign-in under their own policies.
  • Withdrawals are stablecoin transfers on public blockchains (Ethereum or BSC): the destination address and amounts are publicly visible on-chain, permanently. Sending a transfer also discloses it to the blockchain node operator we send it through.
  • We run on Amazon Web Services. Uploaded session content is stored in an Amazon S3 bucket in the ap-east-1 (Hong Kong) region. Our account emails are sent through Amazon SES in the us-east-1 (United States) region. Your data is therefore transferred across borders, and it may be held in a country whose data protection law differs from the one where you live.

Retention and deletion

  • On the My data page you can read every session we hold and delete any session whose reward you have not claimed yet; sessions already paid for can be removed via a manual request so accounts can be settled first.
  • What deletion removes: the uploaded content of that session, both from the database and from the object store.
  • What deletion leaves behind, on purpose: the row recording that the session existed, its token counts, and the hashes and signatures computed from it. Those are what stop the same content being uploaded and paid for a second time. The upload records described above and the reward-ledger entries are also kept.
  • What deletion cannot reach: copies already delivered to a customer, and anything already published as an open dataset. §5 of the Data Collection and Use Authorization states this in full.
  • Backups. We take database backups for disaster recovery and rotate them. A deletion does not reach a backup that was already taken; it takes effect on that copy as the backups holding it are rotated out.
  • No fixed retention period, except that sign-up network records are deleted after 90 days. We keep everything else for as long as it is needed for the purposes stated here. That includes uploads rejected as out of scope: they earn nothing and are never delivered, and there is no scheduled deletion of them. A deletion request that covers them removes them.

Closing your account

Write to support@publicai.io. We handle account closure by hand; there is no self-service button for it. Your sessions and the content you uploaded can be deleted on the terms above. Your authorization record is kept. It is the only proof of what you licensed and on what terms, so §6 of the Data Collection and Use Authorization provides that it outlives the account, and it is used for nothing else.

Your rights

  • Access. You can read your account information and every session we hold from the dashboard. Ask us for anything the dashboard does not show.
  • Correction. Ask us to correct account information that is wrong. The name and the jurisdiction inside an authorization record cannot be corrected in place, because that record cannot be edited after it is written; we can record a correction next to it.
  • Withdrawing consent. You can stop contributing at any time, and you can ask us to delete what we hold. Withdrawing consent to our handling of your personal data does not revoke the license granted in §1 of the Data Collection and Use Authorization, and that license does not limit these rights. §6 of that document says the same thing from the other side.

Email

We send account emails such as verification messages. We do not send per-session acceptance decisions by email — those appear in your dashboard and in the CLI.

Changes

During the beta this policy may change; the current version is always posted at this page with its version number. Material changes will be announced on the site. We do not keep an archive of earlier versions of this page.