AI Usage Policy - Training, retention, third party providers, opt out
Kite AI Usage Policy
Current as of: June 10, 2026
This document describes how Kite uses third-party AI models and services as of the time of writing. Kite may update vendors, model versions, data flows, or feature implementation details over time.
Is Kite an AI video tool?
Kite does not create AI-generated videos and is–at its core–not an AI tool. Kite is a traditional screen recording and editing tool, with some optional AI-powered features.
Currently these AI features include:
- AI voiceovers (text to speech, for narration)
- AI transcription (speech to text, for rendering subtitles)
- Automatic blurring of sensitive information
While many customers get value from the AI features, others use Kite successfully without ever touching them.
As of writing, all Kite AI features are triggered by explicit user actions (e.g. clicking buttons that say "generate" for AI voiceovers, "detect areas to blur", "generate captions", etc).
These features do not run unprompted, and you can turn them off if you want (see below).
We will continue adding new (optional) AI features over time, and will update this document and our release notes as we release them.
Does Kite use your data to train AI models?
Kite does not train or fine-tune AI models on customer data.
Kite does not train AI models itself.
Kite uses third-party AI providers to power specific features, including voiceovers, transcription, and auto-blurs.
Based on Kite’s current providers and the providers’ published policies, customer content sent to these providers by Kite is not used to train or fine-tune their AI models (see table below).
How long are AI inputs and outputs retained?
Kite does not explicitly store your inputs and outputs on our servers any longer than they are needed for processing.
However in most cases, the inputs and outputs of AI features are inherently retained in the Kite project files:
- In AI voiceovers the input script and output audio files are both inherently part of the project file
- Transcription: the original microphone track is already part of the project file, and the subtitles produced by the transcription
- AI auto blurs: the original video frames are already part of the project file
Where Kite project files are stored depends on your Kite plan:
- Pro plan: Kite project files are local files on your computer
- Team plan, projects start as local files and can optionally be sent to the cloud to share with your team
Kite also uses analytics, crash reporting, and session recording tools to understand product behavior, diagnose bugs, investigate errors, and improve reliability.
AI-related prompts or inputs visible in the Kite interface may be captured by these systems when analytics is enabled.
See the "Opting out of usage analytics" section below.
Third party retention policies for AI features are summarized in the table below.
Third party AI provider policies: training and retention
Feature | Provider / models | Data sent | Model training / fine-tuning | Retention |
|---|---|---|---|---|
AI Voiceover | OpenAI | Text submitted for voiceover generation | No–OpenAI states that data sent to the OpenAI API is not used to train or improve OpenAI models unless the customer explicitly opts in. The OpenAI API endpoint table lists speech endpoints as “No” for training. See OpenAI API data controls. | OpenAI lists speech endpoints as having 30-day abuse-monitoring retention and no application-state retention. See OpenAI API data controls. |
AI Voiceover | ElevenLabs | Text submitted for voiceover generation | No–Kite has opted out of ElevenLabs model-improvement training. ElevenLabs states that once the “Improve the models for everyone” toggle is disabled, new submitted data will not be used to train its models. See ElevenLabs data use for model improvement. | ElevenLabs states that it retains data to enhance services, troubleshoot issues, and ensure security. ElevenLabs also states that deleted database items may remain in backups for up to 30 days. See ElevenLabs Zero Retention Mode. |
Transcription | OpenAI Whisper | Microphone audio submitted for transcription in the captions tab | No–OpenAI states that data sent to the OpenAI API is not used to train or improve OpenAI models unless the customer explicitly opts in. The OpenAI API endpoint table lists transcriptions as “No” for training. See OpenAI API data controls. | OpenAI lists transcriptions as having no abuse-monitoring retention and no application-state retention. See OpenAI API data controls. |
AI auto-blurs / UI detection | Google Vertex AI / Gemini | Video frames + optional prompts, explicitly sent to detect UI regions that may need blur treatment | No–Google states that it will not use customer data to train or fine-tune AI/ML models without the customer’s prior permission or instruction. See Google Data Governance | Google states that it does not store prompts or generated output outside the customer’s account longer than reasonably necessary to generate output, absent customer permission or instruction. Google may log prompts for abuse monitoring if suspicious activity is detected; those logs are not used for training and may be stored for up to 90 days. See Google Cloud Service Specific Terms. |
Opting out of usage analytics
Users can disable analytics and crash reporting during setup or later in Kite settings. Disabling analytics and crash reporting may make it harder for Kite’s support and engineering teams to debug issues. See Kite’s help article on Privacy / Offline Modes — Disabling Analytics and Cloud Features.
Opting out of AI and cloud features
Users can also disable Kite cloud and AI features. This keeps project data on-device but disables features that require sending data off-device, including AI voiceovers, shareable links, shared projects, shared templates, AI auto-blurs, and related cloud features. See Kite’s help article on Privacy / Offline Modes — Disabling Analytics and Cloud Features.
AI training data - diversity, representation, and legality
Kite is not an AI company. We do not train AI models.
The responsibility for ensuring, diverse, representative, and legally obtained training data falls to the third party AI providers (Google, Open AI, Eleven Labs, etc).
Updated on: 11/06/2026
Thank you!