Which AI plans have a DPA, a BAA and no training on your chats?
Business and paid developer plans come with a DPA. Consumer plans, free or paid, do not, and none of them has a BAA. Most consumer plans also train on your chats unless you opt out. Pick a plan below to see its DPA, BAA, training default, retention, admin access and EU data residency, each linked to the page it comes from and dated.
- Plans
- 22, from ChatGPT, Claude, Gemini, Microsoft Copilot and Perplexity
- Answers
- DPA, BAA, training, retention, admin access, EU residency
- Checked
- 2 Nov 2026 to 18 Nov 2026
- Sources
- The vendors' own pages, linked in every answer
- Cost
- Free, no sign-up
Check a plan
Pick a tool and a plan. Add a second plan to compare them side by side.
| ChatGPT Plus | ChatGPT BusinessFormerly ChatGPT Team. | |
|---|---|---|
| DPA | ||
| DPADoes the plan come with a data processing agreement (GDPR Article 28)? | No Consumer plans are a contract between you and OpenAI, with no data processing agreement. Checked . Source: Is ChatGPT GDPR compliant? (our review) | Yes OpenAI offers its data processing addendum on Business. Checked . Source: Is ChatGPT GDPR compliant? (our review) |
| BAA | ||
| BAACan a HIPAA covered entity sign a business associate agreement for this plan? | No OpenAI ties its BAA to ChatGPT for Healthcare, ChatGPT for Clinicians, Regulated workspaces and eligible API services. Individual plans are not among them. Checked . Source: OpenAI HIPAA configuration guide | Not documented OpenAI's HIPAA guide names ChatGPT for Healthcare, ChatGPT for Clinicians and Regulated workspaces, not Business. Ask OpenAI before any PHI goes in. Checked . Source: OpenAI HIPAA configuration guide |
| Trains on chats by default | ||
| Trains on chats by defaultAre prompts and chats used to train models unless you turn it off? | Yes, unless you opt out Turn off "Improve the model for everyone" in Settings > Data Controls. Temporary Chats are never used for training. Checked . Sources: Which AI tools train on your data? (our review); Does ChatGPT save your prompts? (our review) | No OpenAI says Business includes no training on your business data by default. Checked . Source: ChatGPT plans (OpenAI docs) |
| Default retention | ||
| Default retentionHow long are chats kept by default? | Until you delete it Chats stay in your history. Deleted chats and Temporary Chats are removed within 30 days, unless OpenAI must keep them for abuse monitoring or by law. Checked . Source: Does ChatGPT save your prompts? (our review) | Until deleted, then 30 days OpenAI lists custom retention controls only for Enterprise and Edu. Our November 2 review found deleted Business chats are removed within 30 days. Checked . Sources: ChatGPT plans (OpenAI docs); Does ChatGPT save your prompts? (our review) |
| What an admin can see | ||
| What an admin can seeCan an employer or workspace admin see or export chats? | No admin Your employer has no admin access to a personal ChatGPT account. A managed work computer can still record what you type. Checked . Source: Can my employer see my ChatGPT prompts? (our review) | Yes Workspace admins can view, export and delete members' conversations, according to OpenAI. Checked . Sources: Can my employer see my ChatGPT prompts? (our review); Does ChatGPT save your prompts? (our review) |
| EU data residency | ||
| EU data residencyCan the data be stored and processed in the EU? | No OpenAI offers no EU data residency option on its consumer plans. Checked . Source: Is ChatGPT GDPR compliant? (our review) | Not documented OpenAI's plan page lists data residency controls under Enterprise and Edu, not Business. Checked . Source: ChatGPT plans (OpenAI docs) |
Every plan, side by side
22 plans from 5 vendors. Each answer links to the page it comes from and shows the date we checked it, between 2 November 2026 and 18 November 2026. This is not legal advice, and vendors change these terms often: confirm with the vendor before you rely on a value.
ChatGPT by OpenAI
Scroll the table sideways to see every plan.
| Free, Plus and Pro | BusinessFormerly ChatGPT Team. | EnterpriseEdu has the same data controls. | APIThe OpenAI API Platform. | |
|---|---|---|---|---|
| DPA | No Consumer plans are a contract between you and OpenAI, with no data processing agreement. Checked . Source: Is ChatGPT GDPR compliant? (our review) | Yes OpenAI offers its data processing addendum on Business. Checked . Source: Is ChatGPT GDPR compliant? (our review) | Yes OpenAI offers its data processing addendum on Enterprise and Edu. Checked . Source: Is ChatGPT GDPR compliant? (our review) | Yes The OpenAI API is covered by OpenAI's data processing addendum. Checked . Source: Is ChatGPT GDPR compliant? (our review) |
| BAA | No OpenAI ties its BAA to ChatGPT for Healthcare, ChatGPT for Clinicians, Regulated workspaces and eligible API services. Individual plans are not among them. Checked . Source: OpenAI HIPAA configuration guide | Not documented OpenAI's HIPAA guide names ChatGPT for Healthcare, ChatGPT for Clinicians and Regulated workspaces, not Business. Ask OpenAI before any PHI goes in. Checked . Source: OpenAI HIPAA configuration guide | Yes, on healthcare workspaces With a signed BAA, OpenAI handles PHI for ChatGPT for Healthcare, ChatGPT for Clinicians and Regulated workspaces. Confirm your workspace is one of them. Checked . Source: OpenAI HIPAA configuration guide | Yes Available through OpenAI's Business Associate and Healthcare Addendum, for eligible endpoints only. Checked . Sources: OpenAI API: your data; OpenAI HIPAA configuration guide |
| Trains on chats by default | Yes, unless you opt out Turn off "Improve the model for everyone" in Settings > Data Controls. Temporary Chats are never used for training. Checked . Sources: Which AI tools train on your data? (our review); Does ChatGPT save your prompts? (our review) | No OpenAI says Business includes no training on your business data by default. Checked . Source: ChatGPT plans (OpenAI docs) | No OpenAI says ChatGPT Enterprise includes no training on business data by default. Checked . Sources: ChatGPT Work admin FAQ; ChatGPT plans (OpenAI docs) | No OpenAI says data sent to the API is not used to train or improve its models. Checked . Source: OpenAI API: your data |
| Default retention | Until you delete it Chats stay in your history. Deleted chats and Temporary Chats are removed within 30 days, unless OpenAI must keep them for abuse monitoring or by law. Checked . Source: Does ChatGPT save your prompts? (our review) | Until deleted, then 30 days OpenAI lists custom retention controls only for Enterprise and Edu. Our November 2 review found deleted Business chats are removed within 30 days. Checked . Sources: ChatGPT plans (OpenAI docs); Does ChatGPT save your prompts? (our review) | Set by your admins Enterprise and Edu include data retention controls for admins. Our November 2 review found deleted chats are removed within 30 days. Checked . Sources: ChatGPT plans (OpenAI docs); Does ChatGPT save your prompts? (our review) | 30 days Abuse-monitoring logs are kept for up to 30 days by default. Zero data retention needs OpenAI's prior approval. Checked . Source: OpenAI API: your data |
| What an admin can see | No admin Your employer has no admin access to a personal ChatGPT account. A managed work computer can still record what you type. Checked . Source: Can my employer see my ChatGPT prompts? (our review) | Yes Workspace admins can view, export and delete members' conversations, according to OpenAI. Checked . Sources: Can my employer see my ChatGPT prompts? (our review); Does ChatGPT save your prompts? (our review) | Yes, Compliance API Enterprise and Edu conversation logs are available to admins through the Compliance API, along with audit logs and usage monitoring. Checked . Sources: ChatGPT Work admin FAQ; ChatGPT Compliance API; ChatGPT plans (OpenAI docs) | Not applicable There is no chat history for an admin to read. Your own application decides what is stored and who sees it. Checked . Source: OpenAI API: your data |
| EU data residency | No OpenAI offers no EU data residency option on its consumer plans. Checked . Source: Is ChatGPT GDPR compliant? (our review) | Not documented OpenAI's plan page lists data residency controls under Enterprise and Edu, not Business. Checked . Source: ChatGPT plans (OpenAI docs) | Yes Enterprise and Edu include data residency controls. Our November 2 review found EU residency covers stored content; inference can run elsewhere. Checked . Sources: ChatGPT plans (OpenAI docs); Is ChatGPT GDPR compliant? (our review) | Yes Projects can store and process data in Europe (EEA and Switzerland) through eu.api.openai.com, for supported models and endpoints. Checked . Source: OpenAI API: data residency |
Claude by Anthropic
Scroll the table sideways to see every plan.
| Free, Pro and MaxUltra is on the same consumer terms. | Team | Enterprise | APIAnthropic's own API, not Claude on AWS or Google Cloud. | |
|---|---|---|---|---|
| DPA | No Anthropic's DPA is part of its Commercial Terms, which cover Claude for Work and the API. Consumer plans run on the consumer terms. Checked . Source: Anthropic: view and sign the DPA | Yes Anthropic's DPA, with Standard Contractual Clauses, is incorporated into its Commercial Terms, which cover Team. Checked . Source: Anthropic: view and sign the DPA | Yes Anthropic's DPA, with Standard Contractual Clauses, is incorporated into its Commercial Terms, which cover Enterprise. Checked . Source: Anthropic: view and sign the DPA | Yes Anthropic's DPA is incorporated into its Commercial Terms, which cover API use. Checked . Source: Anthropic: view and sign the DPA |
| BAA | No Anthropic's plan comparison lists no HIPAA-ready offering on Free, Pro, Max or Ultra. Checked . Source: Claude plans and pricing | No Anthropic's plan comparison lists the HIPAA-ready offering on Enterprise, not Team. Checked . Source: Claude plans and pricing | Yes, once activated The Primary Owner must turn on HIPAA in settings and accept the BAA. Some features, such as Cowork in the cloud and beta features, are not covered. Checked . Source: Anthropic: BAAs for commercial customers | Yes HIPAA-ready API organizations can sign the BAA. Not every API feature is covered; Anthropic lists the eligible ones. Checked . Sources: Anthropic: BAAs for commercial customers; Claude API and data retention |
| Trains on chats by default | Yes, unless you opt out Anthropic lists model training as opt-out on consumer plans. You choose at sign-up and can change it in Privacy settings. Incognito chats are never used. Checked . Sources: Claude plans and pricing; Anthropic: updates to consumer terms; Anthropic: model training (consumer) | No Anthropic does not train on inputs or outputs from its commercial products by default. Checked . Sources: Anthropic: model training (commercial); Claude plans and pricing | No Anthropic does not train on inputs or outputs from its commercial products by default. Checked . Sources: Anthropic: model training (commercial); Claude plans and pricing | No Retained API data is never used for model training without your express permission. Checked . Sources: Claude API and data retention; Anthropic: model training (commercial) |
| Default retention | Until you delete it Deleted chats leave Anthropic's systems within 30 days. With model training on, de-identified copies can be kept up to 5 years; flagged chats up to 2 years. Checked . Source: Anthropic: how long it stores your data | Until the member deletes it Conversations are kept until the member deletes them. Deleted chats leave Anthropic's systems within 30 days. Checked . Sources: Claude Team: data retention and deletion; Anthropic: how long it stores organization data | Indefinite unless set Data is kept indefinitely unless owners set a custom retention period, which must be at least 30 days. Checked . Source: Claude Enterprise: custom data retention | 30 days Inputs and outputs are deleted within 30 days. Covered Models require 30-day retention; zero data retention is on request; flagged content up to 2 years. Checked . Sources: Anthropic: how long it stores organization data; Claude API and data retention |
| What an admin can see | No admin An individual plan has no organization admin, so your employer has no route to it through Anthropic. A work computer can still record what you type. Checked . Sources: Claude plans and pricing; Can my employer see my Claude chats? (our review) | Yes, by export The Primary Owner can export the organization's conversation data. Team has no Compliance API or audit logs. Checked . Sources: Claude: export your organization's data; Claude plans and pricing; Can my employer see my Claude chats? (our review) | Yes, Compliance API The Compliance API gives chats, files and projects to compliance reviewers. Audit logs record activity without chat content. Checked . Sources: Claude: access the Compliance API; Claude: access audit logs; Can my employer see my Claude chats? (our review) | Not applicable There is no chat history for an admin to read. Your own application decides what is stored and who sees it. Checked . Source: Claude API and data retention |
| EU data residency | Not documented Anthropic documents no EU residency option for its Claude apps. Its API residency controls offer US or global processing only. Checked . Source: Claude data residency | Not documented Anthropic documents no EU residency option for Claude apps. Checked . Sources: Claude data residency; Claude plans and pricing | Not documented Anthropic documents no EU residency option for Claude apps. Checked . Sources: Claude data residency; Claude plans and pricing | No Inference can run in the US only or globally, and stored data is US only. There is no EU option. Checked . Source: Claude data residency |
Gemini by Google
Scroll the table sideways to see every plan.
| Gemini app, personal accountgemini.google.com and the mobile app with a personal Google account. | Gemini in WorkspaceThe Gemini app with a work or school account, and Gemini in Gmail, Docs and Meet. | Gemini API, unpaidFree AI Studio use and the API's free quota. | Gemini API, paidA Cloud project with an active billing account. | Gemini Enterprise Agent PlatformCalled Vertex AI until April 2026. | |
|---|---|---|---|---|---|
| DPA | No The consumer app runs under Google's Gemini Apps Privacy Notice, with no data processing agreement. Checked . Sources: Gemini Apps Privacy Hub; Google Gemini DPA explained (our review) | Yes Gemini in Workspace and the Gemini app on a qualifying edition are governed by the Cloud Data Processing Addendum. Checked . Sources: Generative AI in Workspace Privacy Hub; Google Cloud Data Processing Addendum | No No processor terms apply outside the EEA, Switzerland and the UK. Inside them, Google applies its paid-tier data terms to free use too. Checked . Source: Gemini API Additional Terms | Yes Paid use is processed under Google's Data Processing Addendum for Products Where Google is a Data Processor. Checked . Sources: Gemini API Additional Terms; Google processor terms (DPA) | Yes A Google Cloud service, so the Cloud Data Processing Addendum applies through the Google Cloud terms. Checked . Sources: Google Cloud Data Processing Addendum; Google Gemini DPA explained (our review) |
| BAA | No Google's BAA covers listed Workspace and Cloud services. A personal Google account is not one of them. Checked . Source: Google Cloud HIPAA guide | Yes Google's Cloud BAA lists the Gemini app (excluding Gemini in Chrome) and Gemini in Workspace among covered services. Checked . Source: Google Cloud HIPAA guide | No The Gemini API and AI Studio are not on Google's HIPAA service list, and the terms say not to submit sensitive information to unpaid services. Checked . Sources: Google Cloud HIPAA guide; Gemini API Additional Terms | No The Gemini API and AI Studio are not on Google's HIPAA service list, which says not to use unlisted services with PHI. Checked . Source: Google Cloud HIPAA guide | Yes Gemini Enterprise Agent Platform is on the service list of Google's Cloud BAA. Checked . Source: Google Cloud HIPAA guide |
| Trains on chats by default | No, off by default Google says the settings that let it train models on your chats are off by default. Some chats are still read by human reviewers and kept up to 3 years. Checked . Sources: Gemini: manage your history; Gemini Apps Privacy Hub | No Workspace does not use prompts or content to train models, or have people review chats, without your permission. Checked . Source: Generative AI in Workspace Privacy Hub | Yes Google uses prompts and responses to improve its products, and human reviewers may read them. This does not apply in the EEA, Switzerland or the UK. Checked . Source: Gemini API Additional Terms | No Google does not use paid-tier prompts or responses to improve its products. Checked . Source: Gemini API Additional Terms | No Google won't use your data to train or fine-tune models without your prior permission or instruction. Checked . Source: Agent Platform zero data retention |
| Default retention | 18 months History is on by default for adults and deletes itself after 18 months; 3, 36 months or never are options. With History off, chats are kept about 2 months. Checked . Sources: Gemini: manage your history; Gemini Apps Privacy Hub | 18 months Admins set Gemini app history to 3, 18 or 36 months; 18 is the default. With history off, chats are kept up to 72 hours. Checked . Source: Generative AI in Workspace Privacy Hub | Not documented The terms give no period for product-improvement use. Google's abuse-monitoring page keeps prompts and outputs 55 days. Checked . Sources: Gemini API Additional Terms; Gemini API abuse monitoring | 55 days Google logs prompts and responses for a limited period, only to detect policy violations; its abuse-monitoring page says 55 days. Checked . Sources: Gemini API Additional Terms; Gemini API abuse monitoring | Up to 24 hours in cache Implicit caching keeps data up to 24 hours, and the Interactions API stores sessions unless told not to. Zero retention is possible by switching these off. Checked . Source: Agent Platform zero data retention |
| What an admin can see | No admin A personal Google account has no admin, so your employer has no route to it through Google. A work computer can still record what you type. Checked . Source: Can my employer see my AI chats? (our review) | Yes, through Vault On editions with Vault, admins can search, export, retain and hold Gemini app conversations. Audit logs record Gemini activity. Checked . Sources: Use Vault to search Gemini app; Generative AI in Workspace Privacy Hub | Not applicable There is no chat history for an admin to read. Your own application decides what is stored and who sees it. Checked . Source: Gemini API Additional Terms | Not applicable There is no chat history for an admin to read. Your own application decides what is stored and who sees it. Checked . Source: Gemini API Additional Terms | Not applicable There is no chat app. Request-response logging is off by default; your project's own logs decide what admins see. Checked . Source: Agent Platform zero data retention |
| EU data residency | Not documented Google documents no data residency option for the consumer Gemini app. Checked . Source: Gemini Apps Privacy Hub | Yes, by edition Data Regions can keep Gemini app prompts and responses at rest in Europe from Business Standard up. Processing in use needs Enterprise Data Regions. Checked . Source: Workspace Data Regions coverage | Not documented Google makes no residency commitment for unpaid use. Checked . Source: Gemini API Additional Terms | No Paid-tier data may be stored transiently or cached in any country where Google or its agents have facilities. Checked . Source: Gemini API Additional Terms | Yes, EU endpoint The EU multi-region endpoint keeps model processing in EU member states. Global endpoints give no residency guarantee. Checked . Source: Agent Platform data residency |
Microsoft Copilot by Microsoft
Scroll the table sideways to see every plan.
| Copilot, personal accountThe Microsoft Copilot app with a personal Microsoft account. | Microsoft 365 CopilotAlso Copilot Chat signed in with a work or school (Entra ID) account. | |
|---|---|---|
| DPA | No Personal use is governed by the Microsoft Services Agreement and the Copilot Terms of Use, not the data protection addendum. Checked . Sources: Copilot for individuals: your data; Microsoft Products and Services DPA | Yes Covered by the Microsoft Products and Services Data Protection Addendum and the Product Terms. Checked . Sources: Microsoft Products and Services DPA; Data, privacy and security for Microsoft Copilot |
| BAA | No Microsoft's BAA covers in-scope online services for organizations. Copilot on a personal account is a consumer service. Checked . Sources: Microsoft HIPAA and HITECH offering; Copilot for individuals: your data | Yes Microsoft's BAA comes through the DPA by default for covered entities and business associates, and lists Microsoft Copilot and Copilot Chat as in scope. Checked . Source: Microsoft HIPAA and HITECH offering |
| Trains on chats by default | No, in the new app Microsoft says prompts, responses and files in the Copilot app released August 18, 2026 aren't used to train foundation models. The older app trained unless you opted out. Checked . Sources: Copilot for individuals: activity history; Privacy FAQ for the older Copilot app | No Prompts, responses and data accessed through Microsoft Graph aren't used to train foundation models. Checked . Source: Data, privacy and security for Microsoft Copilot |
| Default retention | No period stated The new app keeps past chats so you can return to them, and you can delete them at any time. The older app's FAQ said 18 months by default. Checked . Sources: Copilot for individuals: privacy controls; Privacy FAQ for the older Copilot app | Set by your admins Chats are stored with your organization's Microsoft 365 content. Purview retention policies and holds decide when they are deleted, even after a user deletes them. Checked . Sources: Data, privacy and security for Microsoft Copilot; Microsoft Purview: retention for Copilot |
| What an admin can see | No admin Your employer has no access through Microsoft. On a managed PC, Purview can capture prompts typed into consumer Copilot in Edge. Checked . Source: Can my employer see my Copilot chats? (our review) | Yes, through Purview Admins can search and export prompts and responses with eDiscovery and Content search, and set retention in Purview. Checked . Sources: Data, privacy and security for Microsoft Copilot; Microsoft Purview: search Copilot data |
| EU data residency | Not documented Microsoft documents no data residency option for Copilot on a personal account. Checked . Source: Copilot for individuals: your data | Yes, EU Data Boundary For EU customers Copilot is an EU Data Boundary service. Anthropic models are currently excluded from the boundary. Checked . Source: Data, privacy and security for Microsoft Copilot |
Perplexity
Scroll the table sideways to see every plan.
| Free and ProMax is on the same consumer terms. | EnterpriseEnterprise Pro and Enterprise Max. | |
|---|---|---|
| DPA | No Perplexity's DPA is incorporated into its Enterprise terms. Free and Pro run on the consumer terms of service. Checked . Sources: Perplexity Enterprise terms; Perplexity terms of service | Yes The Enterprise terms incorporate Perplexity's Data Processing Addendum. Checked . Sources: Perplexity Enterprise terms; Perplexity DPA |
| BAA | No Perplexity documents no BAA for consumer plans. Its only HIPAA clause is in the Enterprise terms. Checked . Sources: Perplexity terms of service; Perplexity Enterprise terms | Not documented The Enterprise terms bar PHI unless a BAA has been signed, but Perplexity publishes no BAA or list of covered features. Ask in writing. Checked . Sources: Perplexity Enterprise terms; Perplexity Enterprise |
| Trains on chats by default | Yes, unless you opt out "AI data retention" is on by default on Free, Pro and Max. Turn it off in Settings. Data already collected cannot be removed. Checked . Source: Perplexity: data collection | No The Enterprise terms bar Perplexity from using customer content to train or improve models. Checked . Sources: Perplexity Enterprise terms; Perplexity: data collection |
| Default retention | While your account is open Perplexity keeps your personal information while the account is active and removes it within 30 days of deleting the account. Checked . Source: Perplexity: how long it keeps your data | Files 7 days; chats set by admins Files attached to sessions are deleted after 7 days. Organizations with 50 or more seats, or an Enterprise Max seat, can set how long sessions are kept. Checked . Source: Perplexity: data retention for Enterprise |
| What an admin can see | No admin An individual account has no organization admin. Audit logs and retention controls are Enterprise features. Checked . Sources: Perplexity: Enterprise audit logs; Perplexity: data retention for Enterprise | Yes, audit logs Audit logs, for 50 or more seats or an Enterprise Max seat, include the first 500 characters of each query and answer. Checked . Source: Perplexity: Enterprise audit logs |
| EU data residency | Not documented Perplexity documents no EU hosting option. Its DPA covers transfers out of the EU with Standard Contractual Clauses. Checked . Source: Perplexity DPA | Not documented Perplexity documents no EU hosting option. Its DPA covers transfers out of the EU with Standard Contractual Clauses. Checked . Source: Perplexity DPA |
What the table shows
No consumer plan has a DPA or a BAA
Free or paid, the consumer plans of ChatGPT, Claude, Gemini, Copilot and Perplexity run on consumer terms. Paying for Plus, Pro or Max buys more use, not a contract.
A BAA covers less than the plan name suggests
Claude Enterprise needs its owner to switch HIPAA on. OpenAI ties its BAA to healthcare workspaces and eligible API endpoints. Google and Microsoft cover listed services only, and the Gemini API is not on Google's list.
Two training defaults changed in 2026
Google says training on personal Gemini app chats is off by default, and Microsoft says the consumer Copilot app it released on August 18 does not train foundation models on your prompts. ChatGPT, Claude and Perplexity consumer plans still train unless you opt out.
On a business plan, your employer can usually read your chats
Through an export on Claude Team, the Compliance API on ChatGPT and Claude Enterprise, Google Vault, Microsoft Purview or Perplexity's audit logs. Which one depends on the plan and edition.
EU residency is an enterprise or API feature
OpenAI's API and Enterprise plan, Gemini in Workspace and Agent Platform, and Microsoft 365 Copilot offer it. Anthropic's own options are US or global, and Perplexity documents none.
How to read it
Not legal advice. Vendors change these terms often, so check the date on a value and confirm it with the vendor before you rely on it.
- Yes or No
- The vendor's own page says so. The cell names the page.
- Not documented
- We found no answer on the vendor's current pages. It does not mean no.
- Not applicable
- Developer APIs keep no chat history for an admin to read.
- Sources
- The vendor's current page, read on the date shown. OpenAI's help center refuses automated requests, so a few ChatGPT values come from our own dated review of it, and the cell says so.
- Dates
- Every value shows when it was checked. Our tests flag anything older than 120 days.
The plan is half of it
A plan decides what the vendor promises to do with your prompt. It does not change what is in the prompt, and it only protects work done in the covered account, not a personal one opened in the next tab.
Whatever plan you pick, Locke, made by Sonomos, masks names, account numbers and other sensitive values on your device before the prompt reaches the AI tools it screens. Detection can miss things, and coverage depends on the app and platform.
Questions about AI plans and privacy
Which AI plans come with a DPA?
Business and paid developer plans: ChatGPT Business and Enterprise and the OpenAI API; Claude Team, Enterprise and the API; Gemini in Workspace, the paid Gemini API and Gemini Enterprise Agent Platform; Microsoft 365 Copilot; and Perplexity Enterprise. No consumer plan in the table has one, whether it is free or paid.
Which AI plans offer a BAA for HIPAA work?
Claude Enterprise, once its Primary Owner turns HIPAA on and accepts the BAA, and the Claude API; OpenAI's healthcare workspaces and eligible OpenAI API endpoints; Gemini in Workspace and Gemini Enterprise Agent Platform under Google's Cloud BAA; and Microsoft 365 Copilot under Microsoft's BAA. Each BAA covers listed features only. The Gemini API is not on Google's list, Perplexity's Enterprise terms bar health data without a signed BAA but it publishes none, and no consumer plan has one.
Which AI plans train on my chats by default?
ChatGPT Free, Plus and Pro, Claude Free, Pro and Max, and Perplexity Free and Pro, unless you opt out. The unpaid Gemini API uses prompts to improve Google's products outside the EEA, Switzerland and the UK. Google says training on personal Gemini app chats is off by default, though some chats are read by human reviewers, and Microsoft says its new consumer Copilot app does not use your prompts to train foundation models. Business and paid developer plans do not train on your data by default.
Which AI plans can keep data in the EU?
OpenAI API projects in its Europe region and ChatGPT Enterprise, Gemini in Workspace through Data Regions, Gemini Enterprise Agent Platform through its EU endpoint, and Microsoft 365 Copilot through the EU Data Boundary. Anthropic's own options are US or global, Perplexity documents none, and no consumer plan offers one.
Can my employer read my chats on a business AI plan?
Usually, yes. ChatGPT Business admins can view and export conversations and Enterprise adds the Compliance API; a Claude Team owner can export them and Claude Enterprise has a Compliance API; Google Vault searches Gemini app chats on Workspace editions that include it; Microsoft Purview searches Microsoft 365 Copilot chats; and Perplexity Enterprise audit logs include the start of each query and answer.
Is a plan with a DPA and a BAA enough for GDPR or HIPAA?
No. The DPA or BAA is the contract the law requires before personal data or health information goes to a processor. You still need a lawful basis or a permitted purpose, the right settings, staff training and records. And the contract only helps when people use the covered account rather than a personal one.
What does "Not documented" mean in this checker?
That we found no answer on the vendor's own current pages. It does not mean no. Ask the vendor in writing, and keep the reply, before you rely on that feature.
How current is this AI plan comparison?
Every value shows the date it was last checked against its source; the current table was checked between 2 November 2026 and 18 November 2026. Our tests flag any value older than 120 days for re-checking. Vendors change these terms often, and this page is not legal advice.