From 593c10cd8a7c93f2349ace20a74900d6f82931da Mon Sep 17 00:00:00 2001 From: "prompt.ac/@jeffrey" Date: Mon, 5 Oct 2026 21:01:10 -0700 Subject: [PATCH] Describe Whistlegraph's single first-use AI permission --- system/public/privacy-policy.html | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/system/public/privacy-policy.html b/system/public/privacy-policy.html index 5adff18f63..32b2b6777c 100644 --- a/system/public/privacy-policy.html +++ b/system/public/privacy-policy.html @@ -59,7 +59,7 @@ AI creation requires your permission. Requests, relevant source and version history, drawings and their stroke timing, generated-preview images used for visual review, and sound measurements needed for a request are sent through Aesthetic Computer to OpenRouter and your selected model provider. Available providers include DeepSeek, Moonshot AI, Alibaba/Qwen, MiniMax, and Z.ai. Eligible personal-model accounts can also use Anthropic or OpenAI through our relay.

- Cloud speech and cloud narration each require separate permission. Whisper cloud speech sends each finished recording through AC to OpenAI for transcription and word timing. It uses 40 braincells per recorded second, drawn from the daily allowance first and then purchased credits. Failed transcriptions are refunded. AC stores billing receipts and an audio digest to prevent duplicate charges, but does not store the recording or transcript for this service. A retry result stays in server memory for up to one minute. Earlier personal builds may also use live OpenAI transcription. Cloud narration sends caption text through Aesthetic Computer to ElevenLabs. Without those permissions, speech recognition and narration use the device when available. Typed input remains available. + One first-use AI permission covers creation, Whisper speech transcription, and captions sent to ElevenLabs when optional Jeffrey narration is selected. Earlier builds ask for these permissions separately. Whisper cloud speech sends each finished recording through AC to OpenAI for transcription and word timing. It uses 40 braincells per recorded second, drawn from the daily allowance first and then purchased credits. Failed transcriptions are refunded. AC stores billing receipts and an audio digest to prevent duplicate charges, but does not store the recording or transcript for this service. A retry result stays in server memory for up to one minute. Earlier personal builds may also use live OpenAI transcription. Cloud narration sends caption text through Aesthetic Computer to ElevenLabs. Without those permissions, speech recognition and narration use the device when available. Typed input remains available.

You can withdraw these permissions in Whistlegraph's AI & privacy settings. This stops future transfers for the disabled feature; it cannot recall data already sent. Account deletion removes the drafts and history we store for your account. Provider processing and retention follow their own policies, including OpenRouter, OpenAI, Anthropic, and ElevenLabs. -- 2.54.0