What is a data processing agreement? And which AI plans come with one

The Morse teamPublished November 12, 20269 min read

A data processing agreement (DPA) is a contract in which a company that handles personal data on your behalf, called a processor, promises to use that data only on your instructions, keep it secure, and delete or return it when the work ends. The EU's GDPR requires one whenever a business hands personal data to a processor, and it sets out what the contract must say (EUR-Lex). As of November 2026, OpenAI, Anthropic, Google, Microsoft, and Mistral all offer a DPA for their business and developer products, and none we found covers your chats on a personal plan like ChatGPT Plus or Claude Pro.

This isn't legal advice. We read Article 28 of the GDPR on EUR-Lex, the UK Information Commissioner's Office (ICO) guidance on controller and processor contracts, and each AI company's own DPA, terms, and help pages on November 12, 2026. Every row in the table below links to the page it came from. The companies deserve credit for publishing their DPAs in full, without a sales call, and Google keeps a public archive of the earlier versions of its own.

Which AI companies offer a DPA, and for which plans?

All five we checked do, for business and developer products only. Read November 12, 2026.

A DPA is something an organization accepts as a customer, so all five attach it to their business terms and none to their consumer terms.

Does ChatGPT have a DPA?

Yes for business plans, and only for website visitors' data on Plus and Pro. OpenAI says it can "execute a Data Processing Addendum (DPA) with customers for their use of ChatGPT Business, ChatGPT Enterprise, and the API" (OpenAI, updated January 8, 2026). Its Services Agreement covers "APIs, ChatGPT Enterprise, ChatGPT Business, ChatGPT for Clinicians," and says it "does not apply to OpenAI services used by consumers or individuals" (OpenAI, effective January 1, 2026). Section 5.3 of that agreement says the DPA "is incorporated by this reference," so a Business or API customer already has it. The DPA itself says that by "using the Services, Customer agrees" (OpenAI, effective January 1, 2026).

The "Execute Data Processing Agreement" button on that page opens a form that asks for your company's legal name, your Organization ID, and who will sign (OpenAI). The form says the agreement is "NOT" for "our consumer services (ChatGPT, DALL-E Labs)."

Guides to this question still say Plus has none. Indee's, dated July 21, 2026, says "The free tier and Plus lack a data processing agreement" (Indee). For chats, that's right. But on July 9, 2026, OpenAI published a ChatGPT Sites DPA that "applies to users of ChatGPT Plus, Pro, and other services for individuals, where your published ChatGPT Site collects Personal Data" (OpenAI). It covers the people who visit or sign up on your site. The same page says it "does not apply to your use of ChatGPT in creating or editing a ChatGPT Site, including to your prompts, files, or other content you provide in your conversations with ChatGPT."

For your chats on a personal plan, OpenAI Ireland is "the controller" if you live in the EEA or Switzerland (OpenAI, updated August 24, 2026). The cheapest chat plan with a DPA is ChatGPT Business, which "requires at least 2 Standard or Premium seats," with Standard at $20 per user per month billed annually (OpenAI). How the API's own retention works is in is the OpenAI API private, and what a Business admin can see is in is ChatGPT Business private.

Does Claude have a DPA?

Yes, on Claude for Work and the API, and you never sign it separately. Anthropic says its DPA "is automatically incorporated into our Commercial Terms of Service" and that "when you accept Anthropic's Commercial Terms of Service, you also accept our DPA" (Anthropic, March 16, 2026). The Commercial Terms say "Services under these Terms are not for consumer use" (Anthropic, effective June 17, 2025).

A company can't get a DPA for Claude Pro by paying for it on a company card. Anthropic's Consumer Terms say the Commercial Terms do "not include Claude.ai or Claude Pro use for individuals or entities" (Anthropic, effective October 8, 2025). Anthropic's DPA says "Customer is the controller and Anthropic is Customer's processor," and it promises breach notice "in any event within 48 hours" (Anthropic, effective February 24, 2025). Anthropic's Service Specific Terms put the Team and Enterprise plans under "Claude for Work" (Anthropic, effective August 31, 2026). The Team plan, the cheapest way into those terms, needs "a minimum of two members," at $20 per member per month billed annually for Standard seats (Anthropic).

Do Gemini and Copilot come with a DPA?

Only through a work account. Google's Cloud Data Processing Addendum, last modified October 30, 2026, replaced what Workspace customers knew as the "Data Processing Amendment" (Google). Google's Workspace services summary lists the "Gemini app" among the Main Core Services (Google Workspace). A Workspace customer only needs to opt in "if your Google Workspace or Cloud Identity agreement does not already incorporate the CDPA" (Google Workspace, updated August 10, 2026). To opt in, a super administrator opens Menu, then Account, then Account settings, then Legal and compliance, and clicks Review and Accept. On Google Cloud, the terms say Google processes customer data "in accordance with the Cloud Data Processing Addendum" (Google Cloud).

Microsoft says Copilot and Copilot Chat, "as used by organizations," fall under its Data Protection Addendum, "with Microsoft acting as a data processor" (Microsoft Learn, updated November 4, 2026). The current DPA was published on November 1, 2026, and Microsoft says that when you subscribe to a product under its Product Terms, the DPA defines "the data processing and security terms" (Microsoft). Microsoft's consumer Copilot terms say they "don't apply to Copilot users who sign in with a work, school, or organizational (Entra) account," so a personal sign-in falls under the consumer terms instead (Microsoft, effective August 18, 2026).

Does a DPA stop an AI company from training on your data?

Not always. Mistral's DPA, effective July 27, 2026, says Mistral may process customer personal data "as Controller" for "Training its artificial intelligence models in accordance with its Privacy Policy, unless Customer is or has opted-out of training" (Mistral). The same clause says a thumbs up or down sends the rated input and output to training. Microsoft's DPA keeps a controller role for itself too, though not for training, covering "business operations incident to providing the Products and Services to Customer" (Microsoft, November 2026 edition).

A DPA says what a processor may do with your data, and these clauses are part of what it permits. Read the section on the parties' roles, and look for any purpose the company keeps for itself. Which companies train on personal accounts by default is in which AI companies train on your data.

What does a DPA have to include under GDPR Article 28?

Article 28(3) requires the contract to set out "the subject-matter and duration of the processing, the nature and purpose of the processing, the type of personal data and categories of data subjects" (EUR-Lex). It then lists eight duties the processor must accept:

• process the data "only on documented instructions from the controller"
• make sure its staff are bound by confidentiality
• take the security measures Article 32 requires
• hire another processor only with the controller's authorization, under the same terms
• help the controller answer people who exercise their rights
• help with security, breach notices, and impact assessments under Articles 32 to 36
• delete or return the data at the end, at the controller's choice
• make available the information needed to show compliance, and allow audits

Article 28(9) says the contract "shall be in writing, including in electronic form," which is why a click-through counts (EUR-Lex). Under Article 28(10), a processor that decides the purposes and means of processing on its own "shall be considered to be a controller" for it (EUR-Lex). Breaking Article 28 can bring fines of up to €10 million or 2% of worldwide annual turnover, whichever is higher, under Article 83(4) (EUR-Lex).

The ICO reads the UK GDPR the same way. It says "every time a controller uses a processor to process personal data, there must be a written contract," and gives a company using "a cloud service to store and analyse its data" as an example (ICO). The ICO notes that this guidance is under review after the Data (Use and Access) Act (ICO). It also points out that "sub-processor" is "not a term taken from the UK GDPR itself" (ICO). Who sits on each AI company's subprocessor list is in AI chatbot subprocessors.

Do you need a DPA to use ChatGPT for client work?

It depends on whose personal data goes in. The GDPR "does not apply to the processing of personal data by a natural person in the course of a purely personal or household activity," according to Recital 18, which limits that to activity "with no connection to a professional or commercial activity" (EUR-Lex). A client's customer list, an applicant's CV, or a patient's notes pasted into a chatbot have that connection. For work like that, a client contract or the law may require a processor contract, and a personal plan doesn't come with one. US health data has its own contract, covered in HIPAA compliant AI chatbots.

Ask your lawyer or your client's data protection officer which applies to you, before you paste anything in.

How Morse keeps your own chats off AI company servers

A DPA governs what a processor may do with personal data it handles for you. On a personal plan, no DPA covers your chats, and the company's consumer terms decide what happens to the copy it keeps. Morse keeps your chat history off its servers. It is stored in your browser, encrypted, and our servers don't keep your messages or log what you write.

Where a client contract or the law requires a DPA, use the service you have a DPA with. For your own questions, and for drafts with the client's names and details taken out, each Morse chat runs in one of three modes.

• Private, for drafts no AI host should keep. Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't. Open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok, run here, and new chats start in Private.
• E2EE, for anything no one should read. Your browser checks the secure enclave's hardware attestation, then encrypts the message so only that enclave can read it, and Morse's servers relay ciphertext they can't read. The enclaves are run by third-party providers, not by Morse, and the protection rests on the enclave hardware and attestation working as designed. E2EE comes with Pro, at $19 a month.
• Anonymous, to use GPT, Claude, or Gemini without the provider knowing it's you. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider still sees the words and may keep them for a limited time to check for abuse, under its own policy, so leave identifying details out.

In every mode, unlocking your history takes two keys: one that never leaves your device and one our server gives you when you sign in. Morse doesn't train AI on your content, sell your data, or show ads (privacy policy). Clearing your browser or deleting your account deletes your chats for good, and we can't recover them, so keep a copy of anything you need.

Try a chat from the Morse home page without an account, or start free with no card.