Copilot privacy policy history: every change Microsoft made, dated

The Morse teamPublished November 12, 202610 min read

Copilot's privacy policy is the Microsoft Privacy Statement, read together with Copilot's terms of use and support pages, and the timeline below dates 19 changes to those rules, from Bing's first chat terms in February 2023 to a December 31, 2026 deadline for old chats. The ones that mattered most were ads, allowed in Copilot's terms from the first version; training on your chats, announced in August 2024 and ended for the new app's foundation models on August 18, 2026; and a March 2026 edit to Copilot's privacy FAQ, which Microsoft's change log doesn't mention, that put uploaded files into training and stopped an opt-out from covering past chats.

We read the current Privacy Statement, its Change history page, the Copilot terms and their archive, the Microsoft Services Agreement and its summaries of changes, and Microsoft's Copilot support pages on November 12, 2026. Microsoft doesn't keep past texts of the Privacy Statement online, so for 14 versions from July 2023 to November 2026 we read copies of Microsoft's own page saved by the Internet Archive's Wayback Machine, and compared the Copilot passages in each with the one before. We did the same for the Copilot privacy FAQ.

Microsoft does publish a summary for every month the Privacy Statement changes, back to 2015, on its Change history page, and it keeps seven past versions of Copilot's terms on an archive page. Six of the 13 apps in our comparison of AI policy changes link to nothing older than their current policy.

When did Microsoft change the Copilot privacy policy?

These are the dated changes we found across the Privacy Statement, Copilot's terms, the Services Agreement, the Copilot privacy FAQ, and Microsoft's announcements. Where Microsoft gives only a month, so do we.

What you can do before that last date is in our guide to saving your Copilot chat history.

Does Copilot train on the files you upload?

In the older Copilot app, yes, since a March 2026 edit to Microsoft's privacy FAQ that we found no notice of. In the new app, no. The Wayback Machine's copy of Microsoft's Copilot privacy FAQ from March 6, 2026 says "Regardless of your privacy settings, we do not train our Copilot generative models on your uploaded files," and lists "The contents of files you upload to Copilot" among the things Microsoft doesn't train on (Microsoft, archived). The copy from March 8 has neither sentence.

In their place, the March 8 copy says the data Microsoft uses for AI training includes "your voice and conversation activity with Copilot, including the images or files you upload" (Microsoft, archived). Only "identifying information in uploaded images or files" stays off the training list.

The opt-out shrank in the same edit. On March 6, the FAQ said opting out "will exclude your past, present, and future conversations from being used for training," and that the change would be "reflected throughout our systems within 30 days." On March 8, opting out "will exclude your future conversations." The current FAQ, read November 12, 2026, still says "future" (Microsoft).

The FAQ carries no date of its own, and the Privacy Statement's Change history page lists only retention and marketing edits for March 2026, in two identical entries. We found no Microsoft notice of the FAQ change. Since August 18, 2026, the FAQ applies only to the older version of the Copilot app. For the new one, Microsoft says "your file contents" aren't used to train foundation models.

Does Copilot train on your chats now?

Not in the new app, but the general clause is still there. Microsoft's page for the app released August 18, 2026 says "Prompts, responses, and your file contents when using the Microsoft Copilot app aren't used to train foundation models" (Microsoft).

The Privacy Statement took a month to follow. Its June 2026 version said, in the Copilot section, "In some markets, this data can help train our AI models in Microsoft Copilot unless you opt out." The September 2026 version deleted that sentence (Microsoft, archived). The Change history entry for September says Microsoft added "more information regarding how Microsoft Copilot uses your data" and links to Copilot resources, and doesn't mention the deletion.

The company-wide clause stayed. As of November 12, 2026, the statement says "we may use your data to develop, train, and fine-tune our AI models, including large language models (LLMs)," and its "Learn more" link goes to the older app's FAQ (Microsoft). That wording grew from "develop and train our AI models," added in August 2024, the same month Microsoft announced it would start training on consumer data. How Copilot's training rules compare with other apps' is in which AI companies train on your data.

When did ads come to Copilot?

From the start. The February 2023 terms for Bing's chat already said the service, "including machine to human conversational interactions, may include advertising," and every later version keeps an advertising line (Microsoft).

The Privacy Statement caught up in stages. In June 2024 it said Copilot uses data "to provide relevant advertising." In June 2026 the Advertising section began naming Copilot itself: "Ads may be shown that relate to the current interaction you are having with us, such as your Copilot conversations (including files you share)" (Microsoft, archived). The sentence just before it, in place since at least August 2023, says Microsoft doesn't use "email, human-to-human chat, video calls or voice mail" to target ads. A chat with Copilot isn't on that list.

The August 18, 2026 terms narrowed who sees ads: "If you use Copilot without a paid Microsoft 365 subscription, Copilot may include advertising" (Microsoft). Switching off personalized ads is covered in is Copilot private.

Do you need a Microsoft account to use Copilot?

Yes, again, since August 18, 2026. Bing's chat began in February 2023 requiring "a current Microsoft account" and a signed-in Microsoft Rewards account, and the same terms said that after the first rollout Microsoft "may not require you to be signed in." The October 24, 2025 terms planned for people who "use Copilot without logging in" (Microsoft, archived).

The August 2026 terms reversed that. Their summary says "We've clarified that you must be logged in with a Microsoft account to use Copilot," and the body says "To use Copilot you must have a Microsoft Account" (Microsoft). The same month, Microsoft gave older-app users until December 31, 2026 to sign in to the new app or lose their chats.

How do the Services Agreement updates affect Copilot?

They set the rules that sit under Copilot's own terms, and they change every September 30. The 2023 version added the AI services section, which still says Microsoft "will process and store your inputs to the service as well as output from the service, for purposes of monitoring for and preventing abusive or harmful uses or outputs" (Microsoft). It also bars you from using the AI services, or data from them, "to create, train, or improve (directly or indirectly) any AI technology."

The 2024 version added a Copilot section and clarified "ownership of content, content credentials, and third-party claims." The 2025 version added usage restrictions, and the 2026 version, published July 30 and in force since September 30, renamed the Copilot section and moved arbitration from the AAA to JAMS (Microsoft). The agreement says you accept it by "continuing to use the Services after being notified of a change." Our comparison of how 13 AI apps change their policies puts Microsoft's notice rules next to the others.

Where can you read old versions of Microsoft's privacy statement?

Only in the Wayback Machine. Microsoft's Change history page gives a summary for each month back to July 2015, but as of November 12, 2026 it links to no past texts, and the statement itself has no version history. Copilot's terms are better kept, with seven past versions on Microsoft's archive page, though that page dates them by month only.

Two other gaps turned up. Microsoft's August 16, 2024 announcement on training now redirects to the Copilot blog's front page, so the post that promised "we will always ask first" survives only in archived copies. And the November 2026 statement went up between November 6 and November 10, going by the Wayback Machine's copies. The November 6 copy still says "Last Updated: September 2026." For how OpenAI and Anthropic keep their own archives, see our OpenAI privacy policy history.

How Morse handles your chats

Most rows in that table changed the rules for chats Microsoft already held, on its servers and under your Microsoft account, and the only way to refuse a new version is to stop using Copilot. On Morse, your chat history is stored in your browser, encrypted, and our servers don't keep your messages or log message content.

Morse doesn't train AI on your content, sell your data, or show ads, and it runs no advertising trackers or third-party analytics. Because we don't store chats, we usually have very little to give authorities. The account holds sign-in details, your plan, your balance, and a usage ledger with no content (Morse privacy policy).

You pick, for each chat, who handles your message on its way to a model:

• Private, so no model host keeps a copy. Every new chat starts here, with Grok and open-weight models like 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.
• Anonymous, for GPT, Claude, or Gemini with no account of yours attached. Our proxy removes your account, email, and IP address and sends the message under Morse's own account. The provider sees the message without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy, so leave out names and details that point to you.
• E2EE, for the chats no one but you should read. Your browser checks the enclave's hardware attestation, then encrypts the message so only that enclave can open it, and our servers relay ciphertext they can't read. The enclaves are run by third-party providers, not by Morse, and this rests on the enclave hardware and attestation working as designed. E2EE chats come with Pro, at $19 a month.

When you sign out, your history locks, and opening it takes two keys, one of which never leaves your device. Clearing your browser or deleting your account removes your chats for good, and Morse can't recover them, so keep a copy of anything you need.

You can send a first message from the Morse home page without an account, or start free with no card and no Microsoft account.