Is DeepL safe? What Free, Pro, and the API keep of your text

The Morse teamPublished November 9, 202610 min read

As of November 9, 2026, DeepL is safe for text you don't mind a company keeping, but only DeepL Pro is private: the free DeepL Translator and DeepL Write process what you paste "for a limited period of time to train and improve our neural networks and algorithms," while DeepL Pro deletes your texts once the translation is delivered and says they "will not be used to improve the quality of our services" (DeepL privacy policy). DeepL's free terms also say "you may not use the Free Services for processing content containing confidential or personal data of any kind" (DeepL).

On November 9, 2026, we read DeepL's privacy policy (last updated October 2026), the terms for its free services, the terms for free DeepL accounts, the DeepL Pro terms, its Pro data security page, its pricing page, its Trust Center, its 2025 DeepL Agent announcement, and the Chrome Web Store listing for its browser extension. DeepL deserves credit for how specific its Pro terms are: they say when staff may open a customer's text, for how long, and who has to approve it. The free tier runs under different rules.

DeepL's privacy, plan by plan

Read November 9, 2026, from DeepL's own pages.

Does DeepL store your texts?

Yes on the free version, and only briefly on Pro. DeepL's privacy policy says it processes free users' texts and translations "for a limited period of time" to train its models, and it never says how long that period is (DeepL). A glossary you build without an account "is only stored locally and is not forwarded to DeepL." Signed in, the same policy says "glossaries may be forwarded to DeepL."

DeepL Pro is built the other way round. The policy says Pro texts and documents "will not be permanently stored" and are deleted "after complete performance of the contractually agreed services," translations included (DeepL). The Pro terms add one exception. When "certain error patterns occur," DeepL may store a text for up to 72 hours, encrypted, and the decryption keys go only to "selected employees" in "a logged process," after DeepL's management approves (DeepL Pro terms).

Some records outlast the text on every plan. For each document you translate, DeepL keeps the file type, language pair, character count, and any errors for 14 days, in a database "that can only be accessed by selected employees" (DeepL). It keeps the full IP address of free users for up to 14 days and of Pro users for the length of the contract.

Does DeepL train on your data?

Yes, on every free product, signed in or not. DeepL's free terms reserve the right to process your texts and documents "to train and improve our neural networks and algorithms," and that "also applies to adjustments and corrections you make" (DeepL). Making a free account doesn't change that. The account terms license your content "for the training of DeepL's neural networks" (DeepL). Neither document describes a setting that turns training off; the way out they describe is a paid plan.

DeepL Pro doesn't train on your texts. The privacy policy says Pro texts "will not be used to improve the quality of our services" (DeepL), and DeepL's pricing page lists "No data training" on its cheapest plan, Individual, which it showed us at €7.49 a month billed annually for Germany (DeepL). Documents you upload to build a Translation Memory "are not used to train or improve our models," and neither is the content behind Adaptive Translation. For how DeepL's default compares with ChatGPT, Claude, and a dozen other apps, see which AI companies train on your data.

People may read some free text. In a section of special provisions for users in South Korea, DeepL's privacy policy says that if you enter personal data in the free Translator or Write "contrary to the terms of use, this data may be checked by a human" (DeepL). The policy's general sections don't describe human review elsewhere.

Is DeepL Pro safe for confidential documents?

It's the only DeepL plan that allows them. The Pro terms make DeepL's Data Processing Agreement, published in its Trust Center, "an integral part" of the contract, and they still bar you from sending anything your own "contractual or statutory confidentiality obligations" forbid (DeepL Pro terms). DeepL's Pro data security page says "Texts are never stored or used for model training without your consent" (DeepL).

If you sign up with a work address, DeepL may share it, and sometimes your usage data, with "the holder of the company email domain" (DeepL). Whether a client's NDA allows any machine translation at all is a separate question, which we cover with the codes and contracts in private AI for translators. This isn't legal advice.

Is the DeepL API safe?

The paid API is; the free API key may keep far more. DeepL's pricing page lists three API plans: Developer, to "Test out the API for free," Growth, and Enterprise API (DeepL). The Pro terms define "DeepL API Developer" as "the API provided by DeepL free of charge," and section 3.3.2 says "DeepL API Developer reserves the right to perpetually store any Content or Processed Content" (DeepL Pro terms). Section 8.3.11 says a free API user "shall not process any personal data with the API," and must indemnify DeepL if they do.

The privacy policy says the free API runs under section 3, the free rules, which describe training for "a limited period of time" (DeepL). The contract is the stricter of the two pages, so plan as if the free API may keep your text for good. Paid API Pro requests fall under section 4 and are deleted once the translation is returned.

Where are DeepL's servers?

In Iceland and Sweden for free users, and in Amazon Web Services regions in the EU, the US, and Japan for business and enterprise customers. DeepL SE is based in Cologne, and its Trust Center says that from January 1, 2026, new contracts and renewals move to a hybrid setup (DeepL Trust Center). DeepL's own data centers in "Iceland/Sweden" continue to be used for "research, model training, and servicing free-tier users." Amazon Web Services "handles data processing and storage for business and enterprise customers," and by default their content "may be processed and stored across any of our AWS regions (currently EU, US, and JP)."

Keeping Pro content in one region is now a paid add-on, Data Residency, which the Trust Center says is "available to sales-assisted customers only and cannot be purchased via the DeepL website" (DeepL Trust Center). The privacy policy, updated in October 2026, still says documents on both Pro and the free Translator are "processed on our infrastructure in the EEA" (DeepL). The Pro terms give DeepL "the right to determine the location of data processing" (DeepL Pro terms). If location matters to your contract, get it from DeepL in writing. We compare where a dozen AI apps keep data in where is AI chat data stored.

Is the DeepL browser extension safe?

It's DeepL's own, and it handles more than the text you select. Its Chrome Web Store listing, read November 9, 2026, shows 4,000,000 users, a "Featured" badge, and version 1.107.0, updated that day, from DeepL in Cologne (Chrome Web Store). Its privacy disclosure there says it handles "Personal communications," "Location," "Web history," "User activity," and "Website content." DeepL's privacy policy has sections for its desktop and mobile apps but none that names the browser extension (DeepL). Our guide to whether AI browser extensions are safe explains what that kind of access allows.

The apps follow the free and Pro split. The privacy policy applies the free rules to the free version of the Windows, macOS, iOS, and Android apps, and Pro's to Pro (DeepL). Dictation and read-aloud in the mobile apps use your phone's own speech service, which may send your words to Apple, Google, or Samsung, "possibly also to the USA," and DeepL says it "has no influence on this processing." The mobile apps also use a unique user ID to track your language settings, how many characters you translate, and how you use the copy button.

What about DeepL Write, DeepL Voice, and DeepL Agent?

DeepL Write follows Translator's rules: free Write trains on your text, and Write Pro, a €6.99 a month add-on billed annually, doesn't (DeepL, DeepL).

DeepL Voice deletes your audio, its transcript, and the translation once the service is done (DeepL). Voice to Voice is the exception: an organization's admin can keep transcripts for "anywhere from 20 seconds to 30 days," and the Speaker match feature processes your voice characteristics for up to 30 seconds. For languages without a DeepL voice, it uses backup voices from ElevenLabs.

DeepL Agent, announced on September 3, 2025, works "by using virtual versions of standard tools, like a keyboard, browser and mouse, so it can act on the user's behalf" (DeepL). DeepL's Trust Center lists "DeepL Agent service history" among content "stored with DeepL until the customer actively deletes it" (DeepL Trust Center).

Is DeepL secure?

DeepL publishes its security evidence. Its Pro data security page links a SOC 2 Type II audit, "Certified for security, availability, and confidentiality," and a penetration test by Cure53 (DeepL). Its Trust Center names C5 and ISO 27001 (DeepL Trust Center), and its pricing page lists "Bring your own key" encryption on Enterprise (DeepL). Those audits cover how DeepL protects text on its servers; whether free text goes to training is set by the terms above.

How Morse keeps your text out of training without a Pro plan

On DeepL, what happens to a sentence depends on whether you pay. The free tier trains on it, the free API may keep it for good, and Pro deletes it on servers whose location DeepL picks. Morse doesn't train AI on your content on any plan, Free included, and it doesn't sell your data or show ads. Your chats are stored in your browser, encrypted, not on Morse's servers, which don't keep your messages or log what you write.

On Morse:

• Training on the free tier. No training on any Morse plan, so there's no paid tier to buy for it and no setting to find.
• Free API text kept for good. Morse's servers keep no copy of your messages, and files you attach are sent with the message and never stored by Morse.
• Servers you can't choose. In E2EE, your browser checks the enclave's proof that it's genuine before anything leaves your device, and only that enclave can read the message.

Pick a mode for each chat:

• Private, for the sentences you'd paste into free DeepL. New chats start in Private, on open-weight models such as DeepSeek, Kimi, and GLM. 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.
• E2EE, for a contract or letter no one else should read. Your browser checks the enclave's hardware attestation, then encrypts the text 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 guarantee rests on the enclave hardware and attestation working as designed. E2EE has no web search or image generation, and comes with Pro, at $19 a month.
• Anonymous, to ask GPT, Claude, or Gemini about a phrase. Morse's proxy removes your account, email, and IP address, and the provider sees the text without knowing it's you. It may keep it for a limited time to check for abuse, under its own policy, so leave client names out.

Your history is locked when you sign out, and unlocking it takes a key that never leaves your device plus one Morse's server gives you at sign-in. Clearing the browser deletes it for good, and Morse can't recover it, so save every finished translation in your own files.

So which DeepL is safe for what?

• A menu, a sign, or a public web page: free DeepL is fine.
• Anything with a name, an address, or a client's details in it: not free DeepL, whose terms forbid it. Use DeepL Pro, or a Morse chat in E2EE.
• An app you're building: keep personal data off the free API key, whose terms reserve perpetual storage, and move to a paid API plan before real users' text goes through it.
• A contract that names a country: ask DeepL in writing where your Pro content is processed, since Data Residency is sold only through its sales team.
• The extension: remember it handles "Website content" and "Web history" on the pages where it runs.

Try a chat from the Morse home page without an account, or start free with no card needed. End-to-end encrypted chats come with Pro, at $19 a month.