Service providers

AI and model routing

Google, OpenAI and Anthropic receive inputs and relevant conversation context to provide AI responses, images or other AI features. Under their commercial API terms, OpenAI and Anthropic do not train on API customer data by default. Google's paid API terms say it does not use inputs and outputs to improve its products. Google's unpaid services have different data-use rules.

OpenRouter receives inputs and any reference media needed for a request and routes them to underlying model providers, including BytePlus/Seedance and HeyGen for supported video features. Each underlying provider has its own API terms for training, logging and retention. OpenRouter offers no-training routing choices, but no single training rule covers all of its providers.

fal receives inputs and reference images or video for media generation. Its API and provider terms govern retention and any model-improvement use. Do not assume that all media services have the same no-training terms as commercial text APIs.

OpenAI API privacy · Anthropic API terms · Google API terms · OpenRouter provider data use · fal terms

Voice, audio and talking video

ElevenLabs receives speech text, audio and uploaded voice samples for speech and voice cloning. Deepgram receives audio for transcription. HeyGen receives the text and reference media needed for talking videos. AssemblyAI receives audio and related text only when you enable that integration. LiveKit carries real-time audio and session identifiers; it may be hosted by us or by the configured service.

Voice and video providers have their own API data-use terms. ElevenLabs and HeyGen offer training choices or opt-outs. Deepgram offers a model-improvement opt-out under its API terms. AssemblyAI's terms and your connected account govern its data use. A voice or media request does not have a universal no-training guarantee across these providers.

ElevenLabs data use · Deepgram API data use · AssemblyAI privacy · HeyGen privacy · LiveKit privacy

Email

Amazon Web Services Simple Email Service (AWS SES) receives email addresses, subjects and message bodies to send account, security and optional marketing emails. A mail service you connect receives messages you ask us to send through that connection.

Storage and content delivery

Cloudflare provides network delivery, security and request routing. Cloudflare R2 stores uploaded and generated files. These services receive the stored files or network metadata needed to perform those roles.

Caching and sessions

Redis software supports caching, queues, rate limits and session state on our infrastructure. Upstash is supported for hosted session caching when configured. Self-hosted Redis is part of our hosting, rather than a separate outside recipient.

Payments

Stripe processes payments, subscriptions and eligible creator payouts. It receives payment, account and transaction information needed for billing, fraud prevention and legal duties. We do not store full card numbers or card security codes.

Telephony

Telnyx and Twilio handle phone verification, calls or SMS, depending on the feature and connection. They receive the phone numbers, message content, call audio and recording information needed for that service. Mobile information and SMS consent are not shared for others' marketing.

Analytics and error reporting

Google Analytics receives page and usage events under our regional cookie rules. Analytics storage is off until consent in the EEA, UK, Switzerland and when country is unknown. Elsewhere it is on until you decline. Denied Consent Mode may send cookieless measurement pings. Advertising signals remain denied. Sentry receives error and request metadata to help us fix faults; we remove sensitive details from error reports where possible.

Hosting

Hetzner hosts our application servers, databases and server backups. It provides the infrastructure on which account data, content and logs are stored.

Research and URL safety

DataForSEO and SerpApi receive search terms, keywords and domain inputs when you request research. Book research may send query text to Serper, Crossref and Wikimedia/Wikipedia. Google Web Risk receives URLs for security checks. These services receive the query or URL needed for the requested task.

Data-use limits and connected services

Delivery, storage, caching, payment, telephony, analytics, security, research and hosting providers process data for the roles above under their service terms. We do not send them data for the purpose of training our AI features. Their own retention and service-improvement rules may differ.

Your Charigent improvement setting controls our optional use of your content. It does not change the terms or choices of a provider account you connect. Google and YouTube Limited Use, Slack restrictions and other platform limits still apply. We use connected-platform data only to authenticate you, carry out your requests, maintain connections and resolve issues, as our Privacy Policy describes.

For questions about a provider or a privacy request, email [email protected].