tertiusWebsite, tester account, and sign-in
When you submit the access form, Tertius processes the email address, selected intent (tester access or release updates), the source label used to understand where the signup came from, basic request information needed to protect the form, and a hidden anti-bot field. For tester access, the service generates a short-lived private installer link and marks that address as eligible for a tester account. The address is also retained for important build notices and the preference you selected.
When you sign in from the desktop, Tertius checks whether the email has tester access and sends a six-digit code by email. Code challenges, short-lived access sessions, refresh sessions, attempt counters, and anti-abuse limits are managed by the service. The protected refresh credential is stored through Windows Credential Manager on the tester’s computer.
Tertius says it will not sell submitted email addresses or use tester access as a general weekly newsletter subscription. Release messages should include a way to unsubscribe. Hosting, email delivery, database, and file-storage providers may process the minimum information required to provide those services.
Live meeting and microphone capture
Capture starts only after you explicitly choose to record and Windows grants the required access. A Quick Meeting can capture your microphone and audio playing through the selected Windows output device as separate sources. Tertius shows a visible recording indicator and source-health status; it does not silently start or join a call as a participant.
During a live meeting, bounded audio chunks are sent through the authenticated Tertius service for OpenAI transcription. Tertius is not an offline speech recognizer. You are responsible for giving any required notice and obtaining consent before recording or processing other people.
Encrypted local recovery
To recover an interrupted meeting, Tertius writes checksummed audio checkpoints to encrypted local storage before or while they are processed. Recovery audio is removed after the meeting is finalized. Abandoned recovery data expires after 24 hours. These temporary checkpoints are not the same as the searchable text records kept in meeting history.
Imported audio and video
The original file remains in the folder you selected. Tertius validates the media and extracts a temporary audio copy on the Windows computer. It sends bounded audio segments through the managed service for transcription and removes temporary files after processing or cleanup. The statement “the source video is not uploaded as a whole” does not mean no audio leaves the device.
Transcript, notes, and outcome processing
When you request enhancement, grounded meeting chat, or another outcome, the Tertius service sends the relevant transcript segments, human notes, and server-controlled instructions to OpenAI. The result may include a clean transcript, enhanced meeting notes, cited actions, a summary, grounded answer, follow-up email draft, or another selected mode. Review OpenAI’s current API data controls and terms for information about its handling of API requests.
Service credentials and usage controls
Testers do not provide an OpenAI API key. Tertius stores its OpenAI credential only on the server; it is not embedded in the Windows executable or returned to the desktop. The service validates requests and applies Redis-backed monthly allowances, request limits, concurrency controls, and usage accounting to each authenticated tester account.
The current free allowance includes 120 transcription minutes per month plus managed outcome processing. Usage records include quantities needed to enforce those limits, such as audio duration, audio bytes, processed characters, request counts, and provider token counts. Limits may change during the beta.
Meeting history, Trail, and settings
Tertius stores meetings, transcript segments, human and generated notes, action items, participants you enter or import, and related metadata in local SQLite storage so they can be searched and reopened. If Trail history is enabled, it also stores dictation or imported-media transcript text, the final outcome, and local metadata. The original imported audio or video is not copied into Trail. Application settings, playbooks, lexicon entries, and related preferences are stored locally. Exported files go wherever you choose to save them.
Choices and deletion
- Delete individual transcript segments, choose a transcript auto-deletion period, and choose whether Trail history is enabled.
- Remove exported files from the location where you saved them.
- Sign out to end the current tester session and remove protected account credentials from the desktop.
- Unsubscribe from release notices using the link in a message.
- Do not record or import material unless you have the right and consent to process it.
Security and limitations
No storage or transmission method is risk-free. The current tester installer is not yet code-signed and Windows may display a SmartScreen warning. The beta is not represented as compliant with any specific regulated-industry standard. Organizations should complete their own legal, security, vendor, and data-retention review before use.
Do not use Tertius for highly sensitive, regulated, or confidential recordings unless the managed Tertius service, OpenAI processing, account controls, and beta security posture fit your obligations.
Policy changes
This summary will change as account behavior, allowances, payments, or service providers change. Material changes should be reflected by the updated date at the top of this page. Continued use after an update is subject to the then-current terms.