Plans, teams & roles
My Team
The admin's page: who's on the team, the policies every member's device runs, and the reports you can ask members for. Three tabs, and the middle one is the one that matters.
Available toTeams Admin
My Team appears in your sidebar under an Admin heading, and only for admins. The header reads Invite people and manage team settings; Manage team on the web opens the same team at my.sonomos.ai in your browser. Seats and billing are in Settings › Account › Plan & billing.
Above every tab, a notice headed Before you rely on this for your team lists what to do first. The one that matters most: force-install the Locke extension on every member's browsers by browser policy. Locke screens web AI sites such as chatgpt.com and claude.ai only through the extension, and a member can remove it unless your policy installs it. The full list of gaps is on Known limitations.
| Tab | For |
|---|---|
| Members | Inviting people by email, removing them, and deciding requests to clear detection history. |
| Policies | Writing team policies and pushing them, checking which revision each member's device applied, and choosing how AI services Locke doesn't recognize are handled. |
| Team reports | Asking members to generate a report on their own device, and seeing whether they did. Your own report for this device. |
Members
Four members and one invitation outstanding, with the Detection history card underneath.
- Invite member — type a work email and choose Invite member. Invitations go by email, one address at a time.
- Seats — how many are in use, pending invitations included.
- Status reads Active or Invited. Your own row is marked You · Admin.
- Select ticks members whose detection history you want to authorize clearing (below). An invited address has no device yet.
| Action | Effect |
|---|---|
| Invite member | Sends an invitation to that address. The row reads Invited until they accept. |
| Send a new invite | For an invitation that was missed or expired. Locke asks first: "Send a new invite? The old link will stop working." |
| Revoke | Withdraws an invitation. "The current invitation link will stop working immediately." |
| Remove | Takes an active member off the team. Locke asks first: "They will lose access to this team. You can invite them again later." |
Roles can't be changed on this page, and there is no way to let a whole email domain join: everyone is invited by address.
Adding someone to the team
1Open My Team. The Members tab lists everyone on the team and every invitation that hasn't been accepted yet.
Read all 4 steps as text
- Open My Team. The Members tab lists everyone on the team and every invitation that hasn't been accepted yet.
- Type their work email in the Invite member card and choose Invite member.
- Their row reads Invited, with Send a new invite and Revoke beside it, until they accept the invitation from their email.
- When they accept and sign in to Locke, the row reads Active. Their device is locked to your team from then on, and every policy pushed to all members applies to it the next time Locke syncs.
Detection history
Detection history stays on each member's device and is never uploaded. Members can't clear it themselves: they ask you for a date range, and nothing is deleted until you approve. The Detection history card under the roster is where those requests land.
- Requests from members list the member, the range, an optional reason (Pasted something by accident, Test data, Personal data or Other) and the state. A waiting request has Approve and Deny. Approving deletes nothing on the spot: their app performs the deletion the next time it runs, only for that range, and the clear is recorded in their reports. Denying leaves their history where it is, and they can ask again.
- Clearing without a request: tick members in the Select column, choose Clear detection history for selected members… and confirm with Authorize clearing. It covers all of their recorded history, and each device deletes it the next time Locke runs there.
- History found missing tells you which days lost history on a member's device without an approved clear — never what was detected. Devices check every few minutes while Locke is running and the member is signed in. It is not a guarantee: Locke can't stop a member deleting files their own computer account owns, and it can't see anything recorded and deleted between two checks.
Locke records whether a request was approved and completed, not what was deleted. The member's own reports list every clear with its range and counts.
Policies
The heart of the page. In the app's words: a policy says which AI apps and sites must be screened, which detection types are always masked, and how strict text screening is. Saving keeps a draft; nothing changes on anyone's device until you push it.
One policy pushed to everyone, one draft that no device runs yet.
- Policies — each with who it applies to, its text strictness, how many types it masks and file types it blocks, and a status: Draft, Pushed, or Pushed · unpushed changes.
- Start from a template, or Blank policy, then Add policy.
- Applied by each member — which team policy revision each person's device has applied.
- AI services Locke doesn't recognize — one setting for the whole team.
Templates come from Sonomos, so the list can change. Today they are:
| Template | Starts with | Text strictness |
|---|---|---|
| Baseline | Personal identifiers, government IDs, payment cards and credentials always masked. | Standard |
| HIPAA-focused | Baseline plus health information: medical terms, record and plan numbers, provider IDs and clinical codes. | Strict |
| Finance | Baseline plus banking and tax identifiers, routing codes and transaction references. | Standard |
| Strict | Every detection type Locke knows, masked. | Strict |
A template only fills in the editor. Everything in it can be changed before you save, and a policy you edit later can be put back with Reset to … values.
| Field | What it does |
|---|---|
| Required AI apps and sites | The AI websites and apps Locke screens on every device the policy applies to. Members' devices are locked by the team, so Locke keeps screening every app and site there; leaving one out takes effect only on your own device. |
| Always-masked detection types | Values of these types are always masked; a team policy never sends a type in the clear. Pick by type, or a whole category at once. |
| Strictness › Text | Relaxed, Standard or Strict: how sure Locke must be that a word is a name, a place or an organization before it masks it. Image and file strictness is coming later. |
| File types | Each readable format is Screen or Block. Images and audio, which Locke can't check yet, are sent unscreened unless you set them to Block. Formats Locke can't screen yet, such as archives and Apple iWork files, are always blocked. Members can't change this. |
| Applies to | All members (including you), or selected members. |
Save as draft keeps your changes without touching any device. To send a policy out, choose Save and push in the editor, or Push… on its row, pick All members (including you) or Selected members, and choose Push now. Members' devices apply it the next time Locke syncs, within about a minute, and their own settings for what it covers stop applying. Delete on a pushed policy takes it off members' devices at their next sync; other policies pushed to them still apply.
A member's device is locked: nothing that would weaken protection can be changed from it. Members get a read-only Team policy page instead of Controls, have no Pause, can't let unchecked requests through, and clear detection history only with your approval. For team members shows exactly what they see.
Applied by each member
Pushing is only half of it; this table shows whether each device has caught up. It lists the newest team policy revision any of each person's devices has applied, and when, against the current revision.
| Status | Means |
|---|---|
| Up to date | Their device has applied the current revision. |
| Waiting for the device to sync | Their device applied an older revision and hasn't picked up the latest push yet. It will the next time Locke there syncs — usually Locke isn't running, or the computer is offline. |
| Not applied yet | No device of theirs has reported applying a team policy. |
Someone who uses Locke on more than one device shows their most recently synced one; another of their devices can still be on an older revision until it next connects. If a status stays behind, ask the member to open Locke while they're online.
AI services Locke doesn't recognize
One setting for the whole team, whatever policies apply: what happens when a member's device sends to an AI service that isn't on Locke's list.
| Choice | Effect |
|---|---|
| Each member decides | Members choose. On their locked devices Locke keeps its standard handling: it screens likely AI services and asks about suspected ones. |
| Screen for everyone | Locke screens it before it is sent, as it does for known AI services. |
| Ask me for everyone | Locke holds it and asks: screen and send, send once without screening, or cancel. Members' locked devices keep Locke's standard handling instead. |
| Block for everyone | Locke stops it from being sent. |
| Allow for everyone | Locke sends it without screening. Members' locked devices keep Locke's standard handling instead. |
Screen and Block are enforced on members' devices once Locke there has your team's latest lock, and members are never offered an unscreened send. Let members add their own exceptions (trusted or blocked hosts) controls personal exceptions, but members' own exceptions don't apply on their locked devices either way. Choose Save AI services setting to apply a change.
Choosing text strictness for a team
Stricter settings catch more borderline names, places and organizations and produce more false alarms — and on a team, the false alarms land on other people. Card numbers, keys, IDs and other structured data are masked the same way at every setting. A workable sequence:
- Start at Standard, Locke's default, unless a template set something else for a reason.
- Use Strict for regulated work where a miss costs more than an interruption. The HIPAA-focused template starts there.
- You can't see members' detections, so the signal is what members tell you. Ask them to report a masked value that blocks real work.
- Change strictness on the policy and push it; every device it applies to picks up the change at its next sync.
Team reports
You never see a member's figures. What this tab gives you is a way to ask members to generate a report on their own device for a regime and date range, and to see whether they did. Your own device's reports are here too, and on the Reports page.
One request waiting, two done. Nothing in the table comes from the reports themselves.
- Ask members for a report — pick members (or All members), a Privacy regime (Comprehensive privacy, CCPA / CPRA, HIPAA, UK GDPR or EU GDPR) and a Date range, then Request reports. A member with a request still open can't be asked again until it closes.
- Requests — member, regime, dates, when you asked, and the status: Waiting (with the date it expires), Done, Expired or Cancelled. Cancel withdraws a waiting request.
- Your own report — the same consolidated report the Reports page produces, for this device only.
Each member you ask sees a banner on their Dashboard and Reports page and generates the report themselves. Done means the member's app reported generating it. Locke never receives the report, so it can't confirm that one was built. A request nobody answers expires after 14 days.
A sane rollout order
- Use Locke on your own device for a few days first. You'll write better policy having felt the false alarms yourself.
- Set up the team: Upgrade to Teams in Settings › Account › Plan & billing opens billing in your browser. Teams starts with a 14-day free trial.
- Force-install the Locke extension on members' browsers by browser policy, so web AI sites are screened and the extension can't be removed.
- Write the policy: start from the template closest to your regulatory exposure, keep every AI app and site your team uses required, check the always-masked types, start text strictness at Standard unless the work is regulated, and decide whether images and audio should be blocked. Save it as a draft.
- Invite a handful of people by email. Push the policy to them and watch Applied by each member until they read Up to date.
- Invite the rest, push to All members, and choose how AI services Locke doesn't recognize are handled. After that, check Applied by each member after each push rather than daily.
What your members will experience is covered in For team members — worth reading before you push your first policy, so you know what they'll see.
