Privacy
Last updated: 15 November 2026.
What we read, and what we share
To protect your company, SecTruck reads the email in every Gmail mailbox you connect, as a monitoring and reporting service for you and your staff. Each message is checked for phishing and fraud, and we use what we learn to improve our filters. We also read your Workspace's list of users (names and email addresses), so you can choose which mailboxes to connect, and, where a setting allows it, Google's sign-in and app-access records for them, to spot unusual sign-ins.
What we learn from one customer's mail protects every customer: we share warning signs between customers, such as bad sender domains, bad links and scam patterns, but never the text of your mail.
We don't sell Gmail data, we don't use it for advertising, and we don't use it to train an AI model, ours or a vendor's.
Scams are rarely aimed at one company; most are broad attacks. So a warning sign found in your mail also protects other customers, and theirs protects you. Examples we learn from that contain mail text stay within your company's account.
Who we are
SecTruck is run by Individual Entrepreneur Kenzhaliev Azret Azamatovich (ИП Кенжалиев Азрет Азаматович), 23/12 Aaly Tokombaev St., Block B, Apt. 37, Bishkek 720060, Kyrgyzstan (ул. Аалы Токомбаева 23/12, Блок Б, кв. 37, г. Бишкек, Кыргызстан, 720060). Privacy questions and requests: azret.kenzhaliev@gmail.com.
What we hold
About your company's account:
- Account holders: email, name, a password hash and an encrypted authenticator secret.
- Sign-in sessions: IP address and browser (user agent).
- Dispatchers (seats): name and mailbox.
- Invitations: the invited email address, and its role.
- Alerts: dispatcher names, domains and forwarding addresses, with no message text.
- Rate limits: hashed IP addresses and emails.
- An audit log of account actions.
- Where a company installs our PC guard: each device's label, system edition and build, the guard's version and the result of its security checks.
- Where a company uses our browser extension: the kind of warning and the hostname of the site it was about.
About your mail:
- Most messages are checked in memory and not kept. For a clean message we keep a daily count, a marker with its Gmail message id and no mail text, and a record of its thread with the domains of the people and links in it and hashed participants.
- For a message we flag: its sender, its domains, its subject (up to 200 characters), why it was flagged and what was done.
- Who writes to you: for each known correspondent, a hash of the address and of the display name, the domain, counts, and the first and last dates we saw them.
- Where the takeover checks are switched on: each mailbox's sign-in baseline (a country, or a hash of an IP address or device), its automatic-reply (vacation) setting, as a hash, with a summary that names domains only, and markers for sign-in events already handled.
- Where the site check is switched on: a job for each mail it checks, holding the facts found on the page a link points to, and no URL, path, subject or mail text.
- Your staff's answers in Gmail ("Real threat", "Not a threat", and a label they remove) with the message's Gmail id and mailbox, so we can learn from them.
- If your company reports a message to us, we keep a sanitized, encrypted copy for 1 month so we can review it.
- For some companies, a copy of recent mail sealed under that company's own key, to test and improve our AI checks, where the company has agreed to it, and what we learn from it: learned examples, company memory (baselines) and the rules we learn from your company's mail.
The mail we read includes mail from people outside your company, the senders.
What we keep from all customers together, with no company, mailbox or message in it:
- Warning signs: the registrable domains that links in any customer's mail pointed to, with the first and last day one was seen; and scam patterns we learned.
- Facts about a domain from public sources: its registration record and its mail servers, kept until 30 days after we last checked.
- Where Google Web Risk is switched on: a hash of each link we asked about and Google's answer, kept briefly (6 hours by default for a clean answer).
Why we use it, and on what ground
We use your data to check your company's mail for phishing and fraud, to label and, where your settings allow it, move what looks dangerous, to tell your staff about it, to run the accounts, and to improve our filters.
Who receives it, and where it is processed
- Anthropic, OpenAI and Google (the Gemini API): the AI check. Where your company has agreed to AI review, mail may be sent whole to Anthropic, OpenAI and Google for their AI to rate its risk: suspicious mail and, where a setting allows it, clean mail too, as a screen or a sample. A message your staff un-label or mark "Real threat" may be read again and sent to a second AI to check it. Masked mail blocks may be sent to Anthropic to draft rules, and sealed copies of a company's mail may be sent to the vendors' models to test our checks. See "How we use AI". Each is on its paid API terms that don't train on our data.
- Google Cloud Pub/Sub: tells us that a mailbox has new mail. What it carries is the mailbox's address and a history number, no mail.
- Google (Gmail, Admin and Reports APIs; Cloud KMS): the source of your mail and records, and the service that holds the keys that protect what we store. Where Google Web Risk is switched on: each link's address, cut so that it carries no query, fragment, login or token-like part.
- Vercel: runs the service (region iad1) and sees what passes through it.
- Neon: holds our database.
- Resend: sends our email, such as sign-in and alert emails. It gets the recipient's address and the email's text.
- Public services we query about a domain: DNS, the domain's registration record (RDAP), and, where switched on, the web page a link points to, fetched by our own server. They see the domain or the cut address, never your mail.
- Where broker lookups are switched on: the US Federal Motor Carrier Safety Administration (FMCSA) gets a broker's MC or USDOT number, and nothing else.
This is the full list today. Your data is processed in the United States: our database is in AWS US East (us-east-1), our keys are in Google Cloud KMS in us-east1, and the service runs on Vercel in iad1 (US East).
How long we keep it
- Message markers (Gmail id): 30 days. Thread records: until 90 days after the thread's last message.
- Details of flagged messages: reduced to the sender's domain and the verdict after 180 days, except a flagged message whose quarantine has not been released, which stays in full until it is.
- Your staff's answers in Gmail: 180 days.
- Reported messages: 1 month.
- Alerts and the audit log: 13 months. PC guard reports: 90 days, except the latest report of each device, which stays. Browser-extension warnings: 180 days.
- Sign-in baselines: 90 days after the sign-in was last seen. Site-check jobs: the page facts 28 days and the job 30 days, by default.
- Known correspondents, the hash of the automatic-reply setting and handled sign-in events: until you disconnect, with no end date before that.
- Sealed copies for the AI test: 12 months. Learned examples, company memory and learned rules: deleted by a daily job once your company is disconnected and its AI-test copies are gone; baselines older than 12 months go whatever happens.
- Warning signs we keep for all customers: domains seen in links, 365 days; facts about a domain, 30 days after the last check; Web Risk hashes, briefly (6 hours by default for a clean answer).
- Daily counts: no end date yet.
- The rest: while you are a customer, or until it expires.
When a company disconnects, the data we kept from its mail (flagged and reported messages, threads, markers, known correspondents, the automatic-reply setting's hash, sign-in baselines and events, and site-check jobs) is deleted once the mail we labelled has been restored, and at the latest after 30 days. What stays: daily counts, with no end date yet; your staff's answers, with the message's Gmail id and mailbox, for up to 180 days; alerts and the audit log, for up to 13 months; and the warning signs we keep for all customers, which hold no company or mailbox. Backups: our database keeps its history for 6 hours, and we keep backup copies of it for 7 to 14 days, so deleted data can persist in them until they expire.
Who can see it
Your company's members, by role, and SecTruck's founders. Founders' views of your company are recorded in the audit log.
Our engineers, and the AI coding and review assistants that help them, may read a specific message when a fix or a review is about that message, and the AI vendors above read what we send them. We don't read your mail otherwise.
Google user data
SecTruck's use and transfer of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.
The Gmail permission we ask for, gmail.modify, is broad: Google's own description lets an app read, label, trash and send mail (not delete it for good). We use it only to read mail and to add and remove labels (a label is ours, or the inbox label, which is how mail is moved out and back), and the code that talks to Gmail has no call that sends, trashes or deletes. We also ask to see the people in your Google Workspace domain, to pick seats (read only), and, where a setting allows it, sign-in activity, to spot unusual sign-ins (read only).
We use this data only to give you the checks described here. We transfer it only to the parties above, to give you the feature, for security, or where the law asks.
The rules above apply to what we derive from Google user data as well as to the data itself.
How we use AI
Where your company has agreed to AI review, we may send mail to Anthropic, OpenAI and Google (the Gemini API), in these ways. Each is only for the mailboxes that consent covers, and each stops when we end it:
- Suspicious live mail, whole and, where a setting allows it, with the earlier messages of its thread, up to a size limit, for the AI to rate how risky it is.
- Clean mail, as a screen that reads mail no other AI call reads, and as a sample of other mail.
- A message your staff un-label or mark "Real threat": we read it again from Gmail (we do not store it) and send it to a second AI, from another vendor than the one that flagged it, to check it.
- Masked blocks of mail, sent to Anthropic to draft rules.
- Where a company has agreed to the AI test: sealed copies of its mail, sent to the vendors' models to test and improve our checks.
The models today are claude-opus-5-5, gpt-6.1-sol and gemini-4-carbon-lt for the review, and claude-haiku-5-5 for the screen; the checks and tests use other models of the same vendors, and we may change them. The AI is asked how risky a message is and answers in a fixed form. We send an attachment's name and type, not its contents. Each vendor is on its paid API terms that don't train on our data.
Where AI review is on for your company, the AI's answer can raise or lower the label on a message. It cannot delete or send mail. The AI can be wrong: it can miss a scam, or flag a good message. Where a company's setting allows it, a DANGER message may be moved out of the inbox until your staff restore it.
We don't sell Gmail data, we don't use it for advertising, and we don't use it to train an AI model, ours or a vendor's.
The vendors keep some logs of their own: OpenAI keeps abuse-monitoring logs for up to 30 days by default, and Google's paid Gemini API logs prompts for a limited period for abuse and legal needs only.
We tell the AI that the message is untrusted data and never to follow instructions inside it, and we accept only an answer in the fixed form. If the AI gives no answer, the label stays what our rules gave.
To ask for a review of a label: use "Not a threat" in Gmail, report a message to us, or write to azret.kenzhaliev@gmail.com.
Your choices
- Turn off the AI check: ask us at azret.kenzhaliev@gmail.com, and we end your company's AI consent. That stops every use of AI on your mail described above, except for a mailbox that has its own separate consent for the AI test; ask us to end that too.
- Disconnect: an owner or admin can disconnect in the app. We restore any mail we moved, then delete what we kept from your mail within 30 days.
- Remove our Gmail access: do this after you disconnect, because once it is gone we can no longer restore mail we moved. In your Google Admin console, go to Security > Access and data control > API controls > Manage domain-wide delegation, and delete the entry for our client ID (the setup steps show it). Removing it does not start the deletion; disconnecting does.
- Delete or correct your data: write to azret.kenzhaliev@gmail.com.
What stays after disconnecting is listed under "How long we keep it".
Keeping it safe
The service is reached over HTTPS only. The copies we seal (a reported message, and the AI test's) are encrypted under a key per company that Google Cloud KMS wraps, and the key for a reported message is destroyed when the company's disconnection finishes, and the key for the AI test's copies by a daily job once the company is disconnected. Where the AI reads mail, we tell it the mail is untrusted data, and the code that talks to Gmail has no call that sends, deletes or trashes mail, so a message cannot make it do so.
If we learn of a known or suspected unauthorized access to Google user data, we contain it, tell the customers affected, and notify Google at security@google.com.