KnowMe
How KnowMe uses what you say
Optional dictation and AI voice send bounded microphone audio to OpenAI only while you use them. Grounded AI voice may also send your saved display name, active Context name and kind, selected current confirmed Facts, accepted Evidence excerpts, accepted open or deferred actions you own, and bounded Talk history. KnowMe does not send draft or rejected knowledge, completed or dismissed actions, storage paths, or raw Source files through the grounding tool. KnowMe does not retain raw voice audio; completed Realtime transcripts may be stored in your Companion session. OpenAI API content is not used to train models unless the organization opts in; KnowMe's optional API data sharing was confirmed off on 22 August 2026. OpenAI may retain Realtime abuse-monitoring logs for up to 30 days. KnowMe has not enabled or claimed Zero Data Retention.
Dictation
When you choose Dictate, KnowMe records one bounded audio turn and sends it through KnowMe's server to OpenAI's audio transcription endpoint using gpt-transcribe. KnowMe does not keep the recording. The returned text is placed in the composer, where you can edit or discard it before Send.
Live AI voice
When you choose Start voice, your browser sends microphone audio over an encrypted WebRTC session to OpenAI's Realtime API using gpt-realtime-2.1. gpt-live-transcribe supplies the written transcript. KnowMe stores completed transcript turns in the active Companion session so you can see what was said; it does not store raw voice audio.
To ground a spoken answer, KnowMe's server may send OpenAI your saved display name, the active Context name and kind, a bounded selection of current confirmed Facts and accepted Evidence, accepted open or deferred actions you own, and recent messages from this Talk session. Actions provide planning orientation but do not prove factual claims. Draft, inferred, conflicted, rejected, superseded, completed, dismissed, and foreign-Context material is excluded. The grounding tool does not send raw Source files or storage paths.
Your control
- The microphone starts only after you choose Dictate or Start voice and grant browser permission.
- You can stop, mute, interrupt, or return to text at any time.
- Voice use is metered against your allowance and may be capped or turned off for your account.
- The model cannot confirm, reject, supersede, delete, accept, complete, or send anything for you. Human acceptance is still required.
Provider data controls
OpenAI states that API content is not used to train its models unless an organization opts in. KnowMe's optional API data sharing was confirmed off on 22 August 2026. Under OpenAI's current data-control table, the audio transcription endpoint has no default abuse-monitoring or application-state retention, while Realtime activity may be retained in abuse-monitoring logs for up to 30 days and has no application-state retention. KnowMe has not established Zero Data Retention and makes no Zero Data Retention or data-residency claim.