Four Providers, Four Training Sentences, Four Switches. The One-Page Data Note That Lets You Take the Client's Transcript

Why this pays

Before a client's spreadsheet, transcript or customer list goes into a model, there is an hour of setup and a page to write, and that page is the thing a client's lawyer asks for before signing. Every provider answers the first question in one sentence; the whole rest of the work is the switches that keep each sentence true.

Three of the four cost nothing on their default settings. The hour is not spent on money; it is spent on the note, and the note is what lets you take the job.

The play, in three lines

1. Copy the four training sentences into one document, in the providers' own words, before the first paste.
2. Set one switch per provider: the endpoint at OpenAI, the model name at Anthropic, the tier and the Logs settings at Gemini, the product preference at Mistral. Screenshot each with the date.
3. Attach the document to the client contract as the data note, with the retention period, the deletion rule and what zero retention would take under each sentence.

The path above walks every step. What follows is the reasoning behind each and the numbers you quote from.

Where the money comes from

• The client who would not otherwise send the data. A one-page note quoting the providers' own terms is the difference between a job you can take and one you cannot.
• Nothing extra from the providers. The default settings at OpenAI, Anthropic and Mistral cost nothing; Gemini's paid tier is prepaid then pay-as-you-go.
• The hour, once. The note is reused for every client after the first, with the screenshots refreshed when a page changes.

The pain you are selling the cure for

The client has read that AI tools train on what you paste, and their contract with their own customers says the data goes nowhere. They do not want a reassurance; they want the sentence, the source and the date. A freelancer who hands them that page before they ask is the one who gets the transcript to work on.

Step one: the training sentence, in each provider's words

• OpenAI's data page: as of March 1, 2023, data sent to the OpenAI API is not used to train or improve OpenAI models unless you explicitly opt in to share data.
• Anthropic's retention page: retained data is never used for model training without your express permission; its commercial terms add that Anthropic may not train models on Customer Content.
• Gemini's pricing page, under every model: Free tier content is used to improve Google's products, Paid tier content is not.
• Mistral's commercial terms: no training on Customer Data or Outputs except when you opted in on a product set to opt-out by default or did not opt out on a product set to opt-in by default, when you provide Feedback, as provided in an Order Form, or when you use Labs or Preview Models.

Switch one: OpenAI

• The data page has a table by endpoint. /v1/chat/completions and /v1/responses: no training, abuse-monitoring retention of 30 days, no application state except listed exceptions, eligible for Zero Data Retention. /v1/conversations, /v1/conversations/items and /v1/chatkit/threads: retained until deleted, not eligible. /v1/agents, /v1/assistants and /v1/threads: application state until deleted, not eligible.
• Client work goes through responses or chat completions. Anything built on conversations or threads holds the client's text until you delete it yourself.
• The log no setting removes on a standard account: abuse-monitoring logs may contain prompts and responses, generated for all API usage, retained up to 30 days unless the law or protection of the service requires longer.
• Eligible customers can have their content excluded through Zero Data Retention or Modified Abuse Monitoring, subject to prior approval and additional requirements; once approved, a Data Retention tab appears under Settings, Organization, Data controls. Under Zero Data Retention the store parameter is always treated as false. Write the 30 days into the note.

Switch two: Anthropic

• Conversation content is not retained by default; only what a feature technically needs is kept.
• The exception is a model choice: Claude Fable 5.1, Claude Mythos 5.1, Claude Fable 5 and Claude Mythos 5 are Covered Models that require 30-day data retention and are not available under zero data retention unless expressly authorized. For a client who needs nothing retained, use a model not on that list.
• The general rule: API inputs and outputs deleted within 30 days of receipt, except under services with longer retention you control, such as the Files API.
• Zero data retention is requested from sales and enabled per organization. It does not cover the Claude Console including the playground, so a client's document pasted into the Workbench is outside it; not Claude Managed Agents, whose transcripts persist until deleted; not consumer plans; and CORS is not supported, so a browser tool must call through your own backend. HIPAA readiness can be enabled from Console, Settings, Privacy.

Switch three: Gemini

• The tier: setting up billing in AI Studio moves the project to Tier 1, typically instantly. That is the row under every model turning from yes to no. The tier compare reads it in full.
• On the paid tier, the Logs and Datasets page: logging is opt-in and produces your own data, separate from abuse-monitoring logs; retained 55 days by default, settable to 7, 14, 28 or 55; logs saved into datasets have no retention period.
• The Generate Content API does not store requests by default (store false); the Interactions API does (store true); both can be set per request or per project.
• A dataset shared with Google is processed under the terms for Unpaid Services and may be used to improve and train Google's models, and Google's own page says not to include personal, sensitive or confidential information in one. For client work, never share a dataset.

Switch four: Mistral

• Two kinds of product, opt-out by default and opt-in by default; the training exception depends on which you use and what you chose. Find the preference for each product in your organization's settings and record it; the terms name the preference but not the page.
• Mistral holds a licence to use Customer Data and Outputs to provide, maintain and optimize the products, which the same clause says excludes model training.
• Labs and Preview Models: by using them you acknowledge that Mistral may use Customer Data and Outputs from them to train its models, and opt-out preferences for other products, including zero data retention, do not apply. Off limits for client work, whatever the toggle says elsewhere.

Voices are data too

• ElevenLabs: a Professional Voice Clone stays only in your account unless you share it, and generated content may be used commercially as long as you hold the necessary intellectual property rights.
• A cloned voice is personal data with a face on it. The consent to clone and the promise not to share belong next to the training sentences; the voiceover page reads the licence side.

Asking for zero retention

• OpenAI's requires approval and additional requirements, and its table lists endpoints that stay ineligible even when it is on. Anthropic's is an agreement per organization, requested from sales, with the exclusions above.
• If the client's contract says zero retention, the request goes in writing before the first token, and the note records the date it was granted and the endpoints and models it covers.

What is deleted when you leave

• OpenAI deletes all Customer Content within thirty days of termination and hands nothing back. Anthropic deletes API inputs and outputs within 30 days of receipt, the Files API aside.
• The copies you kept for the client are the only copies. The drafting and transcription pages say the same about their deliverables.

Suspension is a data event

• OpenAI may limit or suspend access for a breach, without prior notice where necessary. Anthropic may suspend all access over a suspected Usage Policy breach and disclaims liability for the loss that causes. ElevenLabs suspends or terminates for conduct such as fraudulent transactions or submitting misleading or unauthorized content.
• A suspended account is one whose stored data you cannot reach or delete on the client's timetable, which is why nothing on this page relies on going back later to clean up.

What to do today

1. Open the four pages linked in the sources and copy the four training sentences into a document.
2. Under each, write the switch you set and the date: the endpoint and the 30-day log at OpenAI, the model name at Anthropic, the tier and the Logs and Datasets settings at Gemini, the product preference at Mistral.
3. Screenshot each setting into a folder with the same date. That document is the data note, and the next client gets it before they ask.