SafeGora in ChatGPT
The data used by your optional workspace connection, how it is retained, and how you control access.
This notice describes the SafeGora connection in ChatGPT operated by OllieLabs LLC. It supplements the SafeGora Privacy Policy and applies when you choose to connect your existing account and verify access in SafeGora. SafeGora accounts remain subject to the Terms of Service, including the requirement to be at least 18 years old.
Information used and shared
SafeGora uses your local account, workspace and identity-provider identifiers, current role, application consent and verified native sign-in claim to check access. Secure sign-in and account linking take place with our identity provider, WorkOS. Authentication secrets belong in the secure sign-in interface, never in chat.
OpenAI receives the permitted results of the tools you request. These include an opaque account/workspace identifier, workflow and source-record identifiers, operational statuses, dates, counts, advisory readiness or loss-control values, proposal status, and links to continue in SafeGora. A verification link contains a short-lived challenge that grants no access by itself.
Proposal requests include the structured record references, dates, priorities and request identifier needed to prepare and safely retry a proposal. SafeGora does not receive your full ChatGPT conversation through these tools. Names, locations, uploaded narratives and documents, medical and injury records, government identifiers, payment-card data and authentication secrets are excluded from workflow inputs and results.
Purpose and recipients
We process this information to authenticate the connection, enforce your current workspace access, answer your requested workflow lookup, and prepare proposals for review in SafeGora. Required native audit evidence records the connected person and the action. SafeGora's existing authorized reviewers can see native proposals and audit records according to their access rules.
Tool results are shared with OpenAI at your request. WorkOS handles account linking, and SafeGora's existing hosting and security providers support the service as described in the sub-processor list. OpenAI's handling of information it receives is governed by its privacy policy and your ChatGPT settings. Human verification, certification, filing and sending communications take place in SafeGora.
Interactive cards and saved preferences
When interactive features are available for your workspace, cards show permitted record identifiers, statuses, dates, counts, advisory flags and links to SafeGora. Selecting a record can share those permitted details with ChatGPT as context for your next request. Cards do not approve, reject, edit or send records. Restricted information and human decisions stay in SafeGora.
When saved preferences are available, SafeGora stores your chosen response language, advisory note length and, if selected, a permitted default establishment for your connected person and workspace. These preferences do not grant access. Current access is checked again for each protected request. Card selections are temporary interface state. To show proposals created in the current conversation, SafeGora stores a hash of the conversation identifier supplied by ChatGPT with the original proposal's required audit evidence. SafeGora does not store the original conversation identifier or chat text through this feature. The hash follows the proposal audit retention described below; displaying associated proposals still requires current workspace and connection access.
When saved preferences are available, disconnecting this workspace removes them. It continues to retain required connection, proposal and audit history as described below.
Optional proposal forms and event updates
When these features are enabled, proposal forms use bounded structured record references and choices. A form reference expires after a short period and can be used once by the connected person. Expiry or use does not automatically delete the reference or required evidence of the resulting native proposal.
If you request event updates, SafeGora stores the subscription, selected event filters, callback address and expiry. The callback signing secret is encrypted. Updates contain only permitted identifiers, statuses, dates, counts and native links. SafeGora also retains queued update records and delivery receipts for reliable delivery and audit. Subscription expiry or disconnect does not automatically delete those records. Delivery stops when consent or the required native access is no longer valid. These features do not authorize a human approval, certification, filing or communication on your behalf.
Retention
A pending verification challenge is valid for ten minutes. SafeGora stores its hash on the server. Your browser tab keeps the challenge in session storage during sign-in, outside redirect URLs, and removes it when the verification page consumes it or the tab closes. Connection bindings and required native proposal, approval and audit records have no automatic expiry in this integration. Disconnecting marks the binding revoked, immediately deletes its saved preferences and pending challenge hash, and keeps the revoked record so an old authorization cannot reopen access. It does not delete existing proposals or required audit records. Operational log retention follows section 6 of the SafeGora Privacy Policy.
Your controls
Revoke this workspace connection under Settings, Integrations, ChatGPT in SafeGora. You can also remove the connection in ChatGPT. Revocation stops subsequent SafeGora workflow access. It does not erase information already returned to OpenAI or your ChatGPT history; use ChatGPT's controls for that information.
To request access to or deletion of your personal information, use the rights process in section 7 of the SafeGora Privacy Policy or contact SafeGora. Requests remain subject to applicable customer instructions, recordkeeping obligations and legal holds.