Platform OverviewOne evidence trail, from intake to Jira.See how it fits together
See it on your documentsA 30-minute call. Bring a vendor you are reviewing now.Book a demo

Do SaaS vendors train AI on your data? What 39 vendors say

What each vendor says about using customer data to train AI models, and about its AI features, in its own words. We do not score the answers: read each statement and its source, and confirm it in your contract.

39 of 39 profiled vendors · Checked 17 Nov 2026 · Every row links to its source

VendorWhat the vendor saysSource
Airtable logoAirtableAirtable states that neither it nor its third-party AI providers use customer data to train current or future models. Admins can choose which workspaces have Airtable AI enabled; on Business and Enterprise Scale plans providers retain no data, while on Free, Team, Pro and Plus plans they may retain data for up to 30 days for safety moderation.cdn.sanity.io
Amazon Web Services logoAmazon Web ServicesSome AWS AI services may use customer content for service improvement, including model training, and may store it in another Region, unless the customer opts out with an AWS Organizations AI opt-out policy. AWS states that Amazon Bedrock content is not used to improve base models or shared with model providers.docs.aws.amazon.com
Asana logoAsanaAsana's product terms state it does not use or permit third-party AI partners to use Customer Data to train generative AI models used to provide Asana AI, and partners must delete customer data after each query. Admins can turn AI request access on or off per user.asana.com
Atlassian logoAtlassianAtlassian says its third-party LLM providers (OpenAI, Anthropic, Google) work under zero data retention and do not train on Rovo inputs or outputs. Atlassian may use de-identified, aggregated metadata to fine-tune open-source models in its own infrastructure, subject to data contribution settings; Enterprise customers can request Atlassian-hosted LLMs only.atlassian.com
Box logoBoxBox says it will not train AI models on customer content without explicit approval, and customers decide whether AI is applied to their content. Box AI uses models from vendors including Microsoft Azure and Google.box.com
Calendly logoCalendlyCalendly states that it does not use customer data to train AI models and does not allow third-party AI providers (currently OpenAI) to do so. There is no opt-out for its AI features generally, but owners and admins can disable Notetaker for the whole organisation.calendly.com
Canva logoCanvaCanva states that it does not use business, team or enterprise content to improve AI-powered features and does not allow third parties to do so, and the related privacy control is disabled for these accounts. Enterprise admins can enable or disable Canva Assistant and control who can use Magic AI features.canva.com
ClickUp logoClickUpClickUp states that its third-party AI providers are prohibited from training on customer data and retain no customer data. The ClickUp AI terms say customers may disable ClickUp AI in their Workspace.clickup.com
Cloudflare logoCloudflareCloudflare states that it does not use customers' content to train any LLMs and does not use customer content to train ML-based products without consent, though detected threat signals may be used to train its ML models. Workers AI does not use customer content to train models or retain inference inputs and outputs after the request.cloudflare.com
Datadog logoDatadogDatadog's terms let it use Generative AI Service output for quality control and service improvement, and use aggregated or anonymised customer data to develop and improve its products; the Bits AI page states zero data retention for third-party AI providers. Datadog does not state on these pages whether customer data trains its models, whether AI features are on by default, or how to opt out.datadoghq.com
Deel logoDeelDeel's privacy policy states that it does not use users' personal data to train or develop AI models, but may use anonymised and aggregated platform usage data to improve its AI systems. It does not state whether AI features are on by default or how to opt out.deel.com
Docusign logoDocusignDocusign says customer data is used for AI/ML training only with the customer's consent, and that consented data is aggregated and anonymised first.docusign.com
Dropbox logoDropboxDropbox says it will not build generative AI models using customer content without consent, and it publishes an AI transparency resource for Dash.dropbox.com
Figma logoFigmaContent training is on by default for Starter and Professional teams, where admins can turn it off, and off by default on Organization and Enterprise plans; Figma for Government and K-12 Education data is not used. Figma does not permit third-party model providers to train on customer data, and Enterprise admins can turn AI features on or off per organisation or workspace.help.figma.com
Fireflies.ai logoFireflies.aiThe Terms of Service say Fireflies.ai will not use User Content to train, retrain, fine-tune or otherwise improve generative AI models, though it may use derived usage and statistical data to improve the service, and it contractually bars vendors from training on customer data. Its subprocessor list names OpenAI, Anthropic, Groq, Baseten, Perplexity, ElevenLabs and AssemblyAI as LLM, search, audio or transcription providers, and the trust center describes a zero-day retention policy with AI vendors.fireflies.ai
GitHub Enterprise Cloud logoGitHub Enterprise CloudGitHub states it does not use Copilot Business or Copilot Enterprise customer data to train AI models. Since April 2026 it uses interaction data from individual Copilot Free, Pro and Pro+ users for training unless they opt out.docs.github.com
Gong logoGongGong states that customer data is never used to train generative AI models, that its AI is developed in-house, and that outputs are generated within each customer's own data environment.gong.io
Google Workspace logoGoogle WorkspaceGoogle states that Workspace does not use customer data to train models without the customer's prior permission or instruction, and content is not human reviewed for model training outside the domain. Access to the Gemini app is on by default and admins can turn it off.knowledge.workspace.google.com
Grammarly logoGrammarlyFor Business, Pro and Enterprise accounts bought through Grammarly's sales team, the Product Improvement and Training control is off by default, and the Customer Business Agreement bars training generative AI on customer data without prior written consent. It is on by default for Free, Premium and individual Pro users and for multi-user accounts bought online, where users or admins can turn it off; the subprocessor list names Anthropic, OpenAI, Azure, Google and Baseten as generative AI or model hosting providers.support.grammarly.com
HubSpot logoHubSpotHubSpot's AI model training setting is on by default and lets HubSpot train its own models on customer data; Super Admins can turn it off, but only for future data. Third-party AI providers may not train on customer data, and accounts with Sensitive Data enabled are opted out automatically.knowledge.hubspot.com
Intercom logoIntercomIntercom trains its Fin CX models on de-identified Fin conversation logs unless a customer opts out in workspace settings, and customers with a BAA, EU or Australian hosted workspaces, older negotiated contracts or trial accounts are excluded automatically. Third-party AI providers are contractually barred from training on customer data and retain nothing once a response is generated.intercom.com
Jasper logoJasperFor Business customers the SaaS Agreement says Jasper will not use, or let a third party use, customer property to train AI models used by the service, but the Terms of Service for Creator and Pro plans let Jasper use inputs and outputs to develop its own AI models, with EU and UK users able to object by emailing legal@jasper.ai. Jasper says its agreements with third-party model providers prohibit training on customer content, and its sub-processor list names Anthropic, Cohere, Google Gemini and OpenAI as LLM providers.jasper.ai
Loom logoLoomLoom AI is on by default for Business + AI and Enterprise workspaces and workspace admins can turn it off for all members. Loom states that its LLM providers, including OpenAI and Google Gemini, do not retain inputs and outputs or use them to improve their services, and Atlassian's data contribution programme for product and AI improvement does not apply to Loom.support.atlassian.com
Microsoft 365 logoMicrosoft 365Microsoft states that Copilot prompts, responses and data accessed through Microsoft Graph are not used to train foundation models. Optional Anthropic models in Copilot are admin-controlled and currently excluded from the EU Data Boundary.learn.microsoft.com
Miro logoMiroMiro's AI terms say it obtains Enterprise customers' consent before using AI inputs and outputs to improve its AI features, while other plans get a means to object. Direct identifiers are removed when data is used this way.miro.com
monday.com logomonday.commonday.com states that customer data, including AI inputs and outputs, is not used to train AI models and that it does not allow others to do so; it uses LLMs through Microsoft Azure and AWS Bedrock with zero retention. Admins can choose whether AI features are enabled in their account.monday.com
Notion logoNotionNotion says that by default neither it nor its AI providers use customer data to train models. Enterprise workspaces get zero data retention on most models, while on other plans AI providers keep data for up to 30 days.notion.com
Okta Workforce Identity Cloud logoOkta Workforce Identity CloudOkta publishes model cards for its AI features stating whether each model trains on Customer Data; the cards checked (Security Access Review, ThreatInsight) say they do not, though ThreatInsight learns from non-personal login telemetry across all customers. No general opt-out statement was found.okta.com
OpenAI logoOpenAIOpenAI says ChatGPT Enterprise business data, including conversations, files and outputs, is not used to train its models by default. Data sent to the API is not used for training unless the customer opts in.chatgpt.com
Otter.ai logoOtter.aiOtter.ai's privacy policy says it trains its proprietary AI on de-identified audio recordings and transcriptions, and users can separately allow or refuse access to their conversations for training and human review in Account Settings. Enterprise workspaces are opted out of AI model training by default; its subprocessor list names Anthropic and OpenAI and says neither trains on nor stores customer data sent through the API.help.otter.ai
Rippling logoRipplingThe AI Services Terms say Rippling will not let its third-party AI providers use customer data, input or output to train or improve their foundation models, though Rippling may use that data to provide, improve and develop its own services. Rippling's AI page says usage data will not be used to train AI models and that admins can enable or disable AI features.app.rippling.com
Salesforce logoSalesforceSalesforce says it operates a zero data retention policy so third-party LLMs do not retain or train on Customer Data, and that it does not currently use Customer Data to train generative AI models; any future training would require customer opt-in under the Product Terms.salesforce.com
Slack logoSlackSlack says it will not use Customer Data to train generative AI models without the customer's affirmative opt-in, and its generative AI uses third-party LLMs that do not receive Customer Data for training. Customer Data is used by default for non-generative global ML models (e.g. emoji and channel recommendations); owners can opt out by emailing feedback@slack.com.slack.com
Snowflake logoSnowflakeSnowflake states it never uses Customer Data to train models made available to its customer base, and that AI models run inside its security and governance perimeter except as the customer elects.docs.snowflake.com
Stripe logoStripeStripe's privacy policy says it may use personal data to train machine learning models used to provide its services, including fraud protection. No business-user opt-out is described there.stripe.com
Twilio logoTwilioTwilio's AI/ML features must be actively enabled by the customer and are optional. Twilio may use AI/ML content to improve those features for that customer and to train models for abuse and security detection, but not to train its other AI/ML features or third-party base models without the customer agreeing to supplemental terms.twilio.com
Workday logoWorkdayFor Universal MSA customers, Workday enables data contributions for AI features in purchased SKUs by default and may use them to train or enhance those features. Admins can turn contributions off per feature or product line, and past contributions are deleted within 30 days.doc.workday.com
Zendesk logoZendeskZendesk says it uses service data to train generic cross-account models, excluding personal-data fields and attachments, and customers can opt out by contacting support. It says third-party LLM providers such as OpenAI never use service data for training.support.zendesk.com
Zoom logoZoomZoom says it does not use audio, video, chat, screen sharing, attachments or other communications-like customer content to train Zoom's or third-party AI models, and third-party model providers may not train on it. Admins control AI features in the web portal; the default setting is not stated on this page.support.zoom.com
Missing a vendor you review?

Tell us which one. We add profiles every week and email you when yours is live.

Questions about this list

Is a public AI statement enough?

No. Statements change and often differ by plan and feature. Ask the vendor to confirm in writing whether your content trains any model, which features are on by default and which model providers process your data.

Why are some vendors missing?

Their public pages did not say how customer data is used for AI when we checked. That is a question to put to them directly.

Where does this come from?

From each vendor’s own trust center, security, legal and privacy pages, as recorded in its profile. Every row links to the page it rests on, so check the source before you rely on it.

Suggest a vendor

Tell us which vendor you review. We will read its public security record, add a profile and email you when it is live.

We use these details to reply about your request. See our Privacy Policy.