Security and data handling

Where your data lives

• OrgSync runs on an Amazon Web Services EC2 server in the United States (N. California, us-west-1). Each company's search index of shared text is kept on that server, separately from every other company's.
• Accounts, companies and shared knowledge are stored in a Supabase PostgreSQL database in the United States (us-west-1). OrgSync connects to it over TLS and verifies the database's certificate against a pinned certificate authority.
• Company files are stored in a private Amazon S3 bucket in the United States (N. California, us-west-1), encrypted at rest, that only OrgSync's server can read or write. File names, folder access and text read from files are stored in the database. The file bucket is not versioned, so a deleted file leaves no stored copy behind.
• About once a week, and before database changes, we take a recovery point: an encrypted snapshot of the server's data disk in AWS (us-west-1) and an encrypted database archive held by our operator in Canada. Each is deleted within 30 days after it is taken. Automatic daily snapshots of the server's data disk are kept for 5 days.
• OrgSync is run from Canada by Fushimi AI Inc.

Who processes it

• AI model: OpenRouter routes shared text, to build the search index, and onboarding chat messages, to draft a work profile, to Microsoft Azure's US endpoint, or to its EU endpoint only when the US endpoint is unavailable. Both routes require zero data retention and deny data collection.
• Search: Amazon Bedrock (Cohere Embed), in the United States (N. Virginia, us-east-1), turns shared text, the requests AI clients send and the asking person's work profile into search vectors. Amazon states that Bedrock does not store these inputs by default, and does not use them to train models.
• Email: Amazon SES, in the United States (us-west-1), sends sign-up verification links, one welcome email and opt-in weekly review reminders.
• Payments: Stripe handles payments. OrgSync never sees or stores card numbers.

Story generation sends eligible company-readable source text, goals, reviews and work-profile changes to OpenRouter for significance assessment, summarization, grounding verification and personalization. Personalization also sends the person's permitted work profile, responsibilities, role objectives, active goals and category preferences. OrgSync checks access before sending inputs and before returning results. Private drafts and raw discussion are not story-model inputs.

Stories uses OpenRouter's global API pinned to Microsoft Azure's US endpoint, azure/us, with zero data retention required, data collection denied and no provider fallback. This does not guarantee end-to-end United States residency through OpenRouter.

We store derived stories and source references, generation metadata, story preferences, daily selections, seen state, likes and comments with their authors, revision numbers, edit and removal timestamps, moderation metadata and links to reviewed company-record changes. Comments are company-readable discussion. Editing a comment replaces its stored body; removing it blanks that body. We do not keep a full comment edit history.

Private drafts are not sent to OrgSync's AI model or search providers until their author shares them. We do not sell your data, use it for advertising, or use it to train AI models.

Company files and their text are not sent to OrgSync's AI model or search providers. A file's text goes only to the AI app of a member who may open it, in a folder that allows AI apps.

The provider register below lists who receives information, why and where. The Privacy notice at orgsync.co/privacy explains collection and retention. A draft data processing agreement, not yet reviewed by a lawyer, is at orgsync.co/dpa.

When your AI client asks OrgSync for context, the cited excerpts go to that AI client, under your AI provider's terms. OrgSync cannot recall copies an AI client has already received.

Service provider register

These providers support the current service. The first three process company content; the others support payments, optional sign-in, support email or font delivery.

Private drafts are stored with our infrastructure providers. They reach our AI model and embedding providers only after sharing. OrgSync's use of model routes requires zero data retention and denies data collection; this is separate from retention of records in OrgSync.

Legacy storage. A stopped Fly.io server in the United States still holds an older copy of service data from before October 4, 2026. It does not run the current service or receive new company content. Its deletion is pending.

Nothing is shared without your approval

• Nothing is captured in the background. Your AI client sends OrgSync a draft, a short summary and the excerpts you chose, only when you ask it to.
• A draft is private to its author. It is shared with the company only after the author approves that exact version in OrgSync's prompt in their AI client or, on a connection they set to confirm in the chat, by saying yes in the conversation. An unshared draft's text is deleted 14 days after it was created or last edited.
• Only Owners and Decision approvers can change a source's knowledge status or accept a suggested correction, and only with their own answer: in OrgSync's prompt, or with a yes in the chat on a connection set to confirm in the chat. Deleting a source always needs OrgSync's prompt.
• The AI cannot answer OrgSync's prompt for you. Software on your own computer that answers prompts automatically could; OrgSync's own plugin never does. On a connection set to confirm in the chat, your AI relays your yes, so a manipulated AI could act without you. Use OrgSync's prompt where your AI client supports it.
• Your own progress on a request (acknowledging an assignment, starting the work or clearing a blocker you reported) is saved when you tell your AI, with no prompt or yes, and its history records it as saved on your word in OrgSync's own fixed words; it never shares a draft, changes a source or changes anyone else's work, and an Owner can require a yes for it.

Access is decided outside the AI

• Who may see what, what has been shared, and each source's knowledge status are decided by OrgSync's own server and database, never by the AI search engine or a model's output.
• The search engine holds derived copies that OrgSync can rebuild from its own records. Before it returns results, OrgSync checks again that the person still has access and that every source it cites is one they may see.
• The service that runs the search engine is reachable only inside OrgSync's server, with a shared secret.
• Each company file's folder decides who can list, download, add, move or delete it, and whether AI apps may read its text. OrgSync's server checks that access on every request, including every AI app request.

Sign-in and connections

• Each person signs in with their own account, with Google or with an email address and password. A password is stored only as a hash.
• A web session ends after 12 hours without use, and after 7 days in any case.
• Each member has a permission profile: Owner, Decision approver or Member. An invitation link works only for the email address it was made for, and expires after 7 days.
• Each AI app connection has its own access token. OrgSync stores only a SHA-256 digest of each token, never the token itself.
• Removing a member revokes their web sessions and AI app connections at once. A disconnected AI app stops working from its next request.

Removed access stays removed after a restore

Every revocation and deletion, such as a removed member, a revoked connection, a discarded draft, a deleted source or file, a file moved between folders, a change to a folder's access or a company deletion, is written to a separate recovery journal before OrgSync's database changes. The journal is kept outside the database's backups, in Amazon DynamoDB (us-east-1), encrypted. It holds identifiers only, never names or content.

If OrgSync is ever restored from a backup, it replays the journal before it opens again, so removed access and deleted content stay removed. If the replay fails, OrgSync stays closed.

After a restore, a folder whose access changed since the backup is closed to Owners only until an Owner sets it again, and a file moved since the backup is placed in a folder only Owners can open.

Protecting the service

• Connections to orgsync.co use TLS.
• Server disks, snapshots, recovery archives and the recovery journal are encrypted.
• Server logs are designed to record fixed status and error codes, not your content, IP addresses or tokens. They are kept for 30 days.
• Our operator can technically reach stored data while running OrgSync, for example when taking or testing a recovery point. We access content only to operate, secure, recover or support OrgSync, at your request, or when the law requires it.

If a breach of our safeguards creates a real risk of significant harm, we will report it to the Office of the Privacy Commissioner of Canada and tell the affected people and companies as soon as feasible.

Export and deletion

• Any member can export the company's shared knowledge, with its history, as JSON or Markdown from Settings › Data. An export never includes anyone's private drafts.
• A member's ZIP export from Settings › Data includes the company files that member may open, checked with the same folder access as a download. Deleting the company deletes every file, folder and access rule.
• Company exports include story discussion and source references, not every Stories table. They do not include a complete copy of generated stories, preferences, selections, seen state or likes.
• Stories and discussion follow the company-content retention rules. Deleting a parent source blocks access immediately and removes its stories and discussion during cleanup. Permitted earlier-version discussion can remain inspectable while its parent record is retained. Content-free generation and accounting receipts remain separately after content deletion.
• Owners can delete the company's content in Settings › Data, after cancelling the plan's renewal.
• Deleting a source or the company stops access at once. We remove active copies, aiming to finish within 7 days. Backups that contain them expire within 30 days after that.
• A company deletion removes the company's content, work profiles, approvals and feedback. It keeps people's accounts, membership and invitation records, and content-free usage records.
• When paid access ends, you have 30 days to export or restart your plan. After that the company's content is deleted in the same way.
• OrgSync cannot erase copies already exported or delivered to an AI client.

What we do not have yet

• OrgSync has no independent security certification or audit report, such as SOC 2 or ISO 27001.
• OrgSync is provided on a best-effort basis, with no promise of uptime or recovery time.
• Uploaded files are not scanned for viruses. OrgSync checks that each file's contents match its type, and always offers files as downloads rather than opening them on orgsync.co.

Report a vulnerability

If you find a security problem in OrgSync, email support@orgsync.co with what you found and how to reproduce it. Our security.txt file at orgsync.co/.well-known/security.txt gives the same address.

• Do not access, change or keep other people's data. If you come across any, stop and tell us.
• Our terms do not allow testing, scanning or probing OrgSync's security without our written permission. Ask us first.
• Give us reasonable time to fix a problem before you make it public.

We read reports Monday to Friday, 9:00–17:00 Toronto time, on a best-effort basis, without a promised response time. We do not offer a bug bounty.

Contact

Security and privacy questions: support@orgsync.co (Monday to Friday, 9:00–17:00 Toronto time). Based in Halifax, Nova Scotia, Canada.