Last updated: November 7, 2026 Effective date: November 7, 2026
1. Who we are
Guardian is an online service that governs what AI agents are allowed to do. It is sold by Odingard Security LLC ("Odingard", "we", "us" or "our"). Our address is Liberty Hill, Texas. You can reach us about privacy at [email protected].
An "agent" is software that acts for a person or a company, for example by sending an email or changing code. Before an important action runs, Guardian gives one of three answers, based on rules you set:
- Allow: the action runs.
- Ask: the action waits until a person you chose approves it.
- Deny: the action does not run.
No AI model makes these decisions. Your rules do.
This policy explains what information we collect when you use Guardian, how we use it, who we share it with, how long we keep it, and the choices you have.
Business customers. When a company uses Guardian, that company decides which agents, connections, approvers and rules to set up. For the data its people and agents put into Guardian, the company is in control and we handle that data on its behalf. Enterprise customers can agree a separate data processing agreement with us.
2. What we collect
We collect only what Guardian needs to work. Here is each kind of information.
Account and sign-in. Sign-in is handled by an outside identity service, Auth0. We store your email address, the ID that Auth0 gives you, and our own account IDs. We do not store your name in our database. Auth0, and the sign-in data kept in your browser, may include the name on your sign-in account. We never see or store your password.
Agents. For each agent you register: its name, status and creation date. You can also tell us which software and AI model it runs on.
Agent requests and decisions. What each agent asked to do, and Guardian's answer (Allow, Ask or Deny). We keep this as a tamper-evident record. Each entry is linked to the one before it, so any change would show.
Approvals. The exact action an approver was shown, and whether they approved or refused it. For an email, this includes the recipients, the subject line and the names of attachments.
Approvals can include information about other people, such as the people an agent emails. You are responsible for having the right to send that information through Guardian.
Receipts and evidence. A digitally signed summary of each governed action, with the full record behind it. See section 5.
Connections to other services. When you connect a service (for example Gmail, GitHub, Slack or Okta), we store which outside account is connected (for example your Gmail address), the permissions you granted, and any limits you set (such as allowed web addresses or file paths). The secret keys and tokens themselves are kept in Google Cloud Secret Manager, a locked store for secrets. They are not kept in our database or in our logs. Your agents never see them.
Push notification subscriptions. If you turn on approval alerts on a phone or browser, we store that device's delivery address, its encryption keys, and a short description of the device. Alerts carry only an approval reference, a short label and a link. They do not include action details, recipients or amounts.
Approval web address. If you set a web address to be notified when an approval is waiting, we store that address.
Billing. Your plan, its status, the end of your current billing period, and the customer and subscription reference numbers from our payment provider, Stripe. You enter card details on Stripe's own pages. Stripe holds your card details. We do not.
Usage counts. One entry for each billable action: reference numbers for you, the agent and the request, the kind of service, the month, and when the action was verified. These entries hold counts and references, not the content of your actions.
Sign-up milestones and referral source. Steps such as "account created", "first agent registered" and "first connection created". If you arrived through a referral, we may also record where you came from.
Other records you or your agents create. Notes an agent saves for later (these are free text and can contain anything, including personal information), spending budgets and payment-type transaction records, household or team memberships (a member can be marked as an adult or a child), labels for your own systems, alert settings, and a log of billing changes with the reasons given.
Rejected sign-in attempts. When a sign-in is rejected, we record which entry point was used and why it was rejected. We do not record who tried.
Server logs. Our servers, which run on Google Cloud, keep logs of their activity. We do not put access tokens or secret keys in logs.
What we do not collect. We do not store the body of your emails or other content from connected services, beyond the action details described above. We do not store tokens from connected services in our database.
Cookies and similar technologies. We use only what sign-in and the console need to work:
- Sign-in storage. To keep you signed in, the console keeps your sign-in tokens from Auth0 in your browser's local storage, and sets two small cookies (one is a copy for older browsers) that only record that you are signed in. While you sign in, a short-lived record of the sign-in attempt is kept in your browser's session storage. Auth0's sign-in pages set their own cookies to run sign-in and keep it secure.
- App storage. The console installs a service worker that stores the console's own files on your device so it loads quickly. It never stores your data or Guardian's answers.
- Push alerts. If you turn on alerts on a device, your browser creates a push subscription, which we keep so we can send you approval alerts.
We do not use analytics, advertising or cross-site tracking cookies.
Do Not Track and Global Privacy Control. Because we do not use cross-site tracking or advertising cookies and do not sell or share your information for advertising, these signals do not change anything about how Guardian works.
3. How we use your information
We use your information to:
- run Guardian: apply your rules, hold actions for approval, send approval alerts, and carry out actions you approved;
- create and sign receipts, and keep the tamper-evident record of what happened;
- sign you in and keep your account and your agents' credentials secure;
- keep each customer's data separate from every other customer's data;
- bill you, collect sales tax, and screen payments for fraud (through Stripe);
- enforce plan limits and spending limits;
- understand how people sign up and start using Guardian, so we can improve it;
- respond to support requests, investigate problems and abuse, and protect the service;
- meet our legal duties, such as tax records.
We do not sell your information. We do not use it for advertising. We do not use your data to train AI models.
Who at Odingard can see your data. Only authorized personnel, and only when needed to provide support, fix problems, keep the service secure, or meet a legal duty. No outside contractor or AI tool is given access to customer data.
4. Who we share information with
Service providers. We use these companies to run Guardian. They handle information only to provide their service to us.
| Provider | What it does for us |
|---|---|
| Google Cloud | Runs our servers, holds our database, and stores secrets (Secret Manager). |
| Auth0 | Handles sign-in. |
| Cloudflare | Sits in front of our service on the network. |
| Stripe | Takes payments, runs subscriptions, collects sales tax, and screens payments for fraud. |
| Browser push services | Run by the maker of your browser. They deliver approval alerts to devices that turned alerts on. |
| Postmark | Sends our service emails to you, such as usage alerts, price or policy change notices and data retention warnings. It receives your email address and the message. |
Services you connect. When an action you allowed or approved runs, Guardian sends the needed information to the service you connected: Gmail, a web service you added with your own access token (for example GitHub, Stripe, Slack, Zendesk or Okta), or a tool server you registered. That service's own privacy policy then applies.
Your own web address. If you ask Guardian to notify a web address you choose whenever an approval is waiting, we send the notice to that address.
People you share receipts with. You can give a receipt to anyone, for example an auditor. That person can read what the receipt contains, including readable action details such as who an email was sent to. You decide whom to share with.
Legal reasons. We may share information if the law requires it, or to protect the rights, safety and security of our customers, the public or Odingard.
Business changes. If Odingard is involved in a merger, sale or similar change, information may move to the new owner. This policy, or one at least as protective, will continue to apply.
Other customers cannot see your data. The database itself keeps each customer's data separate, not only our application code.
5. Receipts
A receipt is a digitally signed record of one governed action. Anyone can check a receipt with our published public key, without a Guardian account and offline. Any change to a receipt makes the check fail. Checking receipts is free on every plan.
A receipt contains the action as Guardian judged it (type, target, any amount or merchant, the stated purpose and the action's details), the answer, the approval if there was one, which permissions applied, what ran, the result, and the full linked record.
Some content appears only as a "digest". A digest is a short code computed from the content. It proves the content has not changed, but the content cannot be read back from it. For email, the body appears only as a digest. The sender, all recipients (including copied and blind-copied), the subject and the number of attachments appear in readable form. The result of an action is also recorded as a digest.
6. Google user data
This section applies when you connect a Google account (for example Gmail).
Limited Use. Guardian's use and transfer of information received from Google APIs follows the Google API Services User Data Policy, including the Limited Use requirements.
What we ask for. For Gmail, Guardian asks for two permissions:
- Create and send email (
gmail.compose). - Read message metadata (
gmail.metadata), such as headers and labels. This does not include the body of your messages.
How we use it. Only to carry out email actions you approve, and to check that they happened. Every email an agent asks to send is shown to a person for approval before it is sent, and no email is sent without that approval.
What we never do with Google data. We do not sell it. We do not share it except as needed to carry out the actions you approved, to comply with the law, or as part of a merger or sale with the protections described above. We do not use it for advertising. We do not use it to train AI models. People at Odingard do not read it unless you agree, it is needed for security, or the law requires it.
Where it goes. The Google token is kept in Google Cloud Secret Manager. Your agents never see it. We keep only the action details described in sections 2 and 5.
Revoking access. When you disconnect a Google connection in Guardian, we delete the connection, try to destroy the stored token, and try to have Google cancel the grant. If Google's cancellation fails, we still delete the connection; you can also remove access in your Google Account settings at any time.
Current status. Our Google app has not yet finished Google's review. Until it does, only users we add by hand can connect a Google account, and that access expires every 7 days. Google Calendar is not available yet.
7. Where your information is stored
We store your information in the United States, on Google Cloud in the us-central1 region (Iowa). Our service providers may process it in other places under their own safeguards.
8. How long we keep information
Retention periods by plan. These are the periods we have set for receipts and records:
| Plan | Kept for |
|---|---|
| Free and Personal | 90 days |
| Family | 180 days |
| Developer | 1 year |
| Business | 3 years |
| Pilot | 1 year |
| Enterprise | As your contract says |
The period runs from the last entry for a request, not the first.
Automatic deletion is not switched on yet. Today, nothing is deleted on these schedules. Until it is switched on, we keep your records until you delete them or delete your account. Before we switch it on, we will update this policy. After that, we will tell you in the console before records expire, and you can export them first.
What we keep longer:
- Billing usage counts: 7 years, for invoice disputes and tax. They hold counts and references, not the content of your actions.
- Deletion markers: when records are deleted, we keep a small marker (the date, how many items were deleted, and the digests of the deleted record sets). This lets a receipt you exported earlier still be matched to our record. It does not contain the deleted content.
- Legal holds: we may place a hold on an account or on one person, for example during a dispute, investigation or where the law requires it. Only a person at Odingard can place or release it. While a hold applies, we cannot delete the records it covers, including through account deletion: account deletion is refused and nothing is deleted. Hold records are kept.
- Billing settings change log: the change is kept, but your name in it is replaced with "deleted-account".
- Rejected sign-in attempts: kept with no set end date. They contain no identity.
- Server logs: kept for about 30 days, Google Cloud's default.
- Backups: we keep the 7 most recent daily database backups, plus 7 days of change logs for point-in-time recovery (Google Cloud SQL's defaults). Older backups are deleted automatically, so data deleted from the database is gone from backups within about 7 to 8 days.
Stripe keeps its own payment records under its own policies.
9. Your choices and rights
Export ("download my data"). In the console you can download one file with your data: your agents, connections (without secrets), permissions, approvals, decision records and their verification, billing status, alert subscriptions (which push service and device, never the delivery keys), sign-up milestones, usage counts, agent notes, budgets and transaction records, household or team memberships, labels and alert settings. The file never includes secrets, tokens or credential fingerprints.
Disconnect a service. Disconnecting deletes the connection, turns off your agents' permission to use it, and tries to destroy the stored secret.
- For Google, we also try to have Google cancel the grant. If that fails, we still delete the connection; you can also remove access in your Google Account settings.
- For other services (for example an access token you pasted in), we delete our copy, but we do not cancel the token at that service. To fully revoke it, cancel it in that service's own settings.
Delete your account. Deleting your account removes every kind of data we hold about you, in one step. We keep only the billing usage counts (7 years), a short record that the deletion happened, and any data under a legal hold (section 8). While a legal hold applies, deletion is refused. We also keep the billing settings change log with your identity replaced by "deleted-account", and copies in server logs and backups expire on the schedules in section 8.
In the console, type "DELETE" to confirm. We then:
- check for a legal hold. If a legal hold covers your account or you, we cannot delete your account while the hold lasts. We will tell you when you try, and nothing is deleted;
- try to cancel your Google grants;
- destroy every stored secret for your connections. Tokens for other services stay valid at that service until you cancel them there;
- try to cancel your Stripe subscription;
- in one step, delete your agents, connections, permissions, approvals, decision and evidence records, push alert subscriptions, sign-up milestones, agent notes, budgets and transaction records, household or team memberships, alert settings, your billing records with us and your user record.
If a stored secret cannot be destroyed, or if deleting your records fails, none of your records are deleted and you can try again. Secrets already destroyed stay destroyed.
In the log of billing setting changes, we keep the change but replace your identity with "deleted-account".
Deleting your account does not delete your sign-in at Auth0 (your email and login). If you sign in again, you get a new, empty account. To have your sign-in deleted too, email [email protected] and we will delete it.
Account deletion covers data tied to your own sign-in. In a household or team, only your own records and membership are deleted. The household or team, and account-wide settings such as labels, are deleted when the last member leaves. A household that remains keeps an internal ID (not your name or email) showing which member created it or added someone. For full deletion or export of a company account, contact us at [email protected]. If you have questions about export or deletion, email [email protected].
Turn off alerts. You can turn off push alerts on any device at any time, in the console or in your browser's settings.
Other requests. You can ask us to tell you what information we hold about you, give you a copy, correct it, or delete it. Email [email protected]. We may need to confirm your identity first. We will respond within 45 days. If your data is held for a business customer, we may pass your request to that business.
We will not treat you differently for using your privacy rights.
US state privacy notice (including California and Texas)
- We do not sell your personal information.
- We do not share it for targeted (cross-context behavioral) advertising.
- In the past 12 months we have collected: identifiers (such as email address and account IDs), commercial information (plan and billing status), internet or network activity (agent requests, decisions and server logs), device information (for push alerts), and information you or your agents enter (approvals, receipts and notes). Section 2 has the details. Sources: you, your agents, the services you connect, and our service providers. Purposes: section 3. Recipients: section 4.
- We do not use or disclose sensitive personal information to infer characteristics about you. How long we keep each category is in section 8.
- Residents of California, Texas and some other states have the right to know, access, correct, delete and get a portable copy of their information, and to opt out of sale, targeted advertising and certain profiling. We do not do any of those three.
- You can use an authorized agent to make a request for you. We may ask for proof that the agent is allowed to act for you.
- If we deny your request, you can appeal by replying to our answer. We will answer the appeal within 60 days. If you are in Texas and we deny your appeal, you can contact the Texas Attorney General.
Users in the European Union and the United Kingdom
We provide Guardian to customers in the United States. If EU or UK privacy law applies to you, this section applies too.
Legal bases. We use your information because:
- it is needed to provide the service under our contract with you (contract);
- it serves our legitimate interests in securing, improving and supporting Guardian, and in preventing abuse, where those interests are not outweighed by your rights (legitimate interests);
- the law requires it, for example tax records (legal obligation);
- you agreed, for example when you turn on push alerts (consent). You can withdraw consent at any time.
Your rights. You can ask to access, correct or delete your information, to restrict or object to how we use it, and to receive it in a portable form. You can also complain to your local data protection authority.
Transfers. Your information is stored in the United States. Where the law requires it, we use recognized safeguards, such as the European Commission's Standard Contractual Clauses, for transfers.
10. Health information
Guardian is not a HIPAA business associate, and we do not sign business associate agreements. Do not send protected health information through Guardian. We plan to offer business associate agreements in the future. Until one is signed with you, do not send protected health information through Guardian. Our demonstrations of health-related features use made-up (synthetic) data only.
11. How we protect your information
- Fail closed. When Guardian is unsure, or part of it fails, the action does not run.
- Secrets are locked away. Connection tokens and our own keys are kept in Google Cloud Secret Manager. Agents never see connection tokens.
- Encrypted database connections. Our database accepts only encrypted connections and cannot be reached from the public internet.
- Separation between customers. The database itself keeps each customer's data apart.
- Key changes. We have a written procedure for changing every kind of key and password.
- Emergency stop. You can stop all of your agents at once.
- Incident response. We have a written plan for security incidents. If a serious incident affects your data or your approvals, we will tell you what happened, what was affected, what we did, and what you need to do. We will also give any notice the law requires.
No system is perfectly secure, and we cannot promise that information will never be accessed without permission.
12. Children
Guardian is not directed to anyone under 18. We do not knowingly let anyone under 18 create an account.
On household plans, an owner of the household can add family members and mark a member as a child. A child member may be added only by that child's parent or legal guardian, who is responsible for the child's use and for giving any consent needed. Only an adult owner of the household can add or remove members. The adult account holder who creates a household is its first owner. Children can never be owners and can never approve actions.
Household members marked as children must be at least 13. Do not add anyone under 13. We do not knowingly collect personal information from children under 13. If we learn we have, we will delete it.
If you think a child has given us information without a parent's consent, email [email protected] and we will delete it.
13. Changes to this policy
We may update this policy. We will change the "Last updated" date at the top. If a change is material, we will give you at least 30 days' notice before it takes effect, in the console or by email to the address on your account. We will update this policy before we switch on automatic deletion.
14. Governing law
This policy is governed by the laws of the State of Texas, except where the privacy law of the place you live gives you rights that cannot be changed by agreement.
15. Contact us
Odingard Security LLC Liberty Hill, Texas [email protected]