Termsby GlyphThe brief · Mon Nov 9Google updated its Agent Platform documentation to clarify that zero data retention for OpenAI models is disabled by default and requires project approval. Anthropic clarified that organizations with zero data retention arrangements cannot use browser-based API calls due to CORS restrictions. Vercel expanded its zero data retention availability to include teams with purchased credits, while Modal confirmed that container logs retain stdout and stderr data despite zero retention for other endpoints. Leaders should verify if their current API configurations or browser-based integrations conflict with these updated retention and security constraints.
Read more
What to do
- Verify if your Google Agent Platform project requires manual activation for zero data retention.
- Check if your Anthropic API integration relies on browser-based CORS requests while using zero data retention.
- Audit Modal container logs to ensure sensitive request payloads are not written to stdout or stderr.
Says yes
17Does it train on API and business data?Yes, may train on your info“Improve our online operations and the Services, including for the development, training and finetuning of generative AI tools”
Yes, anonymized, opt-out“Fin can use anonymized customer data for model fine-tuning, and you can opt out at any time with data deleted within 30 days.”
Yes, unless ZDR is on“create derivative works thereof as necessary for GRADIUM to analyse and improve the current and future Services, and train and develop any underlying AI model”
Yes, license covers AI training“in order to provide and improve our Services, including for Intercom AI model training”
Yes, inputs and outputs“Jasper uses Customer Property, including Input and Output, to provide, maintain, and improve the Services, including for the purpose of developing and enhancing”
Yes, unless agreed in writing“Unless otherwise expressly agreed in writing, Customer Content may be used for the foregoing purposes.”
Yes, for speculative decoding models“process Input and Output to (a) provide, maintain, monitor, protect, and support the Services, Token Factory Service or (b) train smaller Models used for Speculative Decoding.”
Yes, anonymized and aggregated“We utilize anonymized and aggregated data to train and improve our artificial intelligence (AI) models.”
Yes, by default under the MSA“the Customer Materials and Output to analyze and improve Scale's products and services, including machine learning models.”
Yes, User Content used“otherwise disclose to third parties any User Content for the purpose of providing, developing and enhancing the Services, including training models, and to comply with applicable law.”
Yes, unless you opt out“On all paid Plans (whether self-serve or under an executed Order Form), Customer Content may be used for Model Training”
Yes, unless you opt out“we may use Content to improve and develop our Services (but you can opt-out to prevent us from using your Inputs and Outputs to train our models”
Says no
65No, never trains“Airbyte does not collect or use customer data to develop, improve, or train generalized artificial intelligence and/or machine learning models”
No, unless you consent“Alibaba Cloud will not use your Member Content to develop, train or improve the AI models made available through the Service, unless you separately and expressly provide your consent to us.”
No, partners barred too“Amplitude contractually prohibits our AI partners from using your data to train or improve their models”
No, unless you opt in“By default, we will not use your inputs or outputs from our commercial products (e.g. Claude for Work, Anthropic API, Claude Gov, etc.) to train our models.”
No model training“Anyscale shall not, nor shall it permit any AI Services to, use Customer Property to train, fine-tune, or improve any artificial intelligence or machine learning model”
Not on paid use“we do not use Inputs and Outputs associated with paid instances of our Services to train and improve our Services.”
Never used for training“Neither Atira nor our model providers use customer data to train or improve AI models.”
AWSNo“This means your data is not shared with model providers, and is not used to improve the base models.”
Never trains on your data“Baseten will not use Customer Content, including Customer Models, Model Outputs, or inference logs, to train, fine-tune, or otherwise develop machine learning or artificial intelligence models.”
No, never trains“We will not use Your Personal Data or Customer Content to train or improve AI models”
No, never without approval“Box also does not train AI models based on customer data”
No, OpenAI barred from training“Data sent to OpenAI via Braze is not used to train or improve OpenAI models”
Only in part
15Only de-identified metadata“Atlassian may use metadata to fine-tune open-source models that operate strictly within Atlassian's infrastructure, solely to improve the quality of responses and experiences we deliver to customers”
Only de-identified, aggregated data“This data may be used for both internal and external purposes, such as analyzing, enhancing, testing, training and improve Baz Services (including through the use of artificial intelligence) and for research purposes”
Opt out in settings“You can opt out from your prompts and generations being used to train Cohere models in your dashboard settings at any time.”
Feature models only, not generative“These models may be trained on your documents and metadata, and power features within Dropbox such as improved search relevance”
Off for Org and Enterprise“By default, content training is turned off on Organization and Enterprise plans”
Off for sales-bought plans only“For multi-user team accounts purchased directly through our website, Product Improvement and Training will initially be on, and an admin can opt out for their entire organization.”
Anonymized usage data only“JFrog does not use customer artifacts or binaries to train AI language models”
De-identified usage data“to use Usage Data in de-identified or aggregated form to maintain, improve, train, finetune, and enhance Keenable's current and future products, services and technologies”
Paid API: you can opt out“Customers retain full control over this processing and have the right to opt out at any time.”
Depends on the provider“Wherever possible, OpenRouter works with providers to ensure that prompts will not be trained on, but there are exceptions.”
Only custom models for you“Raindrop may use Your Data to train or fine-tune machine learning or artificial intelligence models solely for your benefit and use”
No for Enterprise only“For Enterprise users, Replit does not use customer content to train its models.”
Pages that changed
12Added 1 paragraph; the clause quoted for retention and subprocessors is no longer on the page.New text: “(b) Company configures its AI model Subprocessors so that they do not retain Customer Data after processing a request.”
Added 2 paragraphs.New text: “Does CodeRabbit only review the diff?”
Rewrote 2 paragraphs.New text: “DuckDB has the official documentation , blog , and library . There are also a few third-party resources which can help you learn more about DuckDB:”
Added 7 paragraphs and removed 5.New text: “This page covers the Claude API ( api.anthropic.com ), Claude Platform on AWS, Claude Models on Google Cloud (Anthropic managed) , and Claude in Microsoft Foundry , where Anthropic is the data processor. On Claude Models on Google Cloud (Go”
Rewrote 2 paragraphs.New text: “The inference_geo parameter is available on the Claude API (first-party), Claude Platform on AWS , and Claude Models on Google Cloud (Anthropic managed) . On Claude Models on Google Cloud (Google managed) and Amazon Bedrock, the inference r”
Removed 2 paragraphs.New text: “Admins can control AI content training settings.”
Rewrote 5 paragraphs; the clause quoted for zero retention is no longer on the page.New text: “In the dbt platform , dbt Wizard is available in the Studio IDE and the dbt Wizard home tab . dbt Copilot is available in Insights .”
Added 13 paragraphs and removed 3.New text: “Shopify : Connecting a Shopify store is covered in Shopify Store Data below”
Added 14 paragraphs and removed 4.New text: “Run the following curl command to turn off all caching for a Google Cloud project. This disables both in-memory caching and durable caching, and no cached data is retained. To run this command, a user must be granted the Agent Platform Admi”
Removed 30 paragraphs and added 5; the clause quoted for residency is no longer on the page.New text: “A Pinecone index lets you store and search your data. A document index stores JSON documents whose fields you declare in a schema at creation. It can combine dense vector, sparse vector, and full-text ( string ) fields in one document, so a”
Rewrote 3 paragraphs.New text: “Last Updated: October 27, 2026 ( previous version )”
Rewrote 1 paragraph; the clause quoted for trains on api data and zero retention is no longer on the page.New text: “These Terms of Service - Enterprise (this " Agreement ") are entered into between SpaceX Super Intelligence LLC (" SpaceXSI ") and the business customer of SpaceXSI for SpaceXSI business services (" Customer "), which Customer may be identi”