Is Microsoft Copilot safe? Age rules, teen protections, privacy and state letters

The short answer

It depends on who is using it and what they share. As of October 29, 2026, Microsoft's Copilot terms say you must be at least 13, or older where local law requires. Microsoft's help center says Copilot doesn't personalize results for young people, doesn't show them personalized ads, and lets parents block Copilot or turn off features such as image generation, voice mode and AI companions.

On privacy, Microsoft's help center says humans may review Copilot conversations for safety, legal, product improvement or troubleshooting reasons. For the updated Copilot app, Microsoft says prompts, responses and file contents aren't used to train foundation models; its terms say it may use your content to operate and improve Copilot, and its pages for the earlier app say chats could be used for AI training unless users opted out. You can delete single chats or your whole Copilot history.

Microsoft was among 13 companies that two groups of state attorneys general wrote to in 2025. In July 2025, Missouri's attorney general's office sent Microsoft questions about an answer its letter attributes to Copilot. These are letters, not findings.

Which Copilot this guide covers

This guide covers Microsoft Copilot used with a personal Microsoft account at copilot.microsoft.com or copilot.com and in the Copilot apps. Microsoft says its consumer Copilot privacy pages do not apply to work or school accounts, and its age-limits page says it does not cover Microsoft 365 Copilot.

Microsoft says an updated version of the Copilot app for web, desktop and mobile became available on August 18, 2026, and that several of its older help articles apply only to the earlier version. Where this guide relies on a page for the earlier version, it says so.

Who can use Copilot and what teens get

• Minimum age: Microsoft's Copilot Terms of Use, effective August 18, 2026, say you must be at least 13, and sometimes older depending on the law where you live. They say users under 18 may have some features turned off or limited for legal or safety reasons.
• Age checks: we did not find a Microsoft page that describes how Copilot determines or verifies a user's age.
• Personalization: Microsoft's Copilot for young people page, updated August 18, 2026, says Copilot doesn't personalize results for young people based on their conversation history, and that it may use data differently or limit certain features for them.
• Ads: Microsoft says young people may see ads in Copilot, but that these are based on what they are asking in the chat, not on other information about them. Its privacy controls page says Copilot doesn't show personalized ads to users under 18, whatever their settings.
• Content: Microsoft says Copilot checks for certain risky or harmful content. Its young people page for the earlier app adds that on sensitive topics such as mental health or substance use, Copilot may give general guidance, point to safer sources or encourage talking to a trusted adult.
• Training (earlier app): Microsoft's young people page for the earlier Copilot app says chats of users under 18 aren't used to train AI models.

Parental controls

As of October 29, 2026, Microsoft's help center says parents or guardians can use Microsoft Family Safety, Google Family Link or Apple Family Sharing to:

• block the Copilot app or Copilot in a browser;
• set limits on when Copilot can be used, such as certain hours;
• choose which features are available, such as image generation, voice mode, group chat or AI companions.

Microsoft says some of these controls may require adding the young person to the parent's family group. We did not find a Microsoft page saying parents can see their child's Copilot conversations.

Microsoft also notes that features such as Groups may let other people see a chat if a user shares a conversation or joins a group.

Content rules and crisis features

Microsoft's Copilot Terms of Use, effective August 18, 2026, include a code of conduct. Among other things, it bars using Copilot to:

• harm yourself or others, or create content that glorifies violence or suicide;
• create or share adult content or child sexual exploitation or abuse material;
• create or share graphic violence, hateful content or terrorist or violent extremist content;
• create or edit images, voice or video of other people without their permission.

The terms say Microsoft may block, restrict or remove prompts or content that break them, and may limit, suspend or permanently end access, unless the law prohibits it.

Crisis features: Microsoft's Transparency Note for Copilot, dated August 18, 2026, says Copilot uses classifiers to detect language suggesting suicide or self-harm intent. It says the system then disengages from the conversation and provides crisis support resources, such as the 988 Suicide and Crisis Lifeline in the U.S., and that Copilot does not provide counseling.

Accuracy: Microsoft's terms say Copilot can make mistakes, should not be used as a substitute for professional advice, and that users should check information before relying on it.

Privacy: training, human review and retention

• Training (updated app): Microsoft's help page on Copilot activity history, dated August 18, 2026, says prompts, responses and file contents in the Microsoft Copilot app aren't used to train foundation models. It says optional feedback may be used to improve Copilot, but not to train the foundation models Copilot uses. Microsoft's terms say it may use your content to operate and improve Copilot.
• Training (earlier app): Microsoft's privacy FAQ for the earlier Copilot app, last updated August 31, 2026, says Microsoft uses data from Copilot and other services for AI training, except for certain users and users who opt out. It lists exclusions including signed-in users under 18, users who are not signed in, and users in some countries, and says Microsoft removes identifiers such as names, phone numbers and device and account IDs before training. Its privacy controls page for that app names two opt-out toggles, Training on conversation activity and Training on voice conversations, under Privacy in Copilot's settings, and says opting out applies to future conversations.
• Opt-out in the updated app: the privacy controls page for the updated app, last updated September 17, 2026, does not list a model-training toggle. Microsoft's Privacy Statement, last updated September 2026, which covers Microsoft products generally, says Microsoft may use data to develop, train and fine-tune its AI models; it does not say whether conversations in the updated Copilot app are used for training.
• Human review: Microsoft's young people page says humans may review Copilot conversations for safety, legal, product improvement or troubleshooting reasons. Microsoft's terms say Copilot may involve automated and manual (human) processing of data and advise users not to share information they don't want reviewed. The privacy FAQ for the earlier app says an opt-out of human review is not available.
• Retention: Microsoft's help page on conversation history, last updated August 13, 2026, before the updated app's release, says Copilot keeps the last 18 months of interactions in your conversation history. The privacy FAQ for the earlier app gives the same 18-month default. We did not find a retention period on the help pages for the updated app.
• Health data: Microsoft's terms say data in Copilot Health is not used to develop or train AI models.

How to delete Copilot chats and memory

As of October 29, 2026, Microsoft's help pages for the updated Copilot app describe these steps:

• One chat: in the Chats list, select More (the three dots) next to the chat, then Delete.
• Your whole history: sign in at account.microsoft.com/privacy, go to Privacy, then Empower your productivity, then Copilot, then Your Copilot app activity history, then Copilot apps. Choose Delete all activity history and confirm with Clear. The same page offers Export all activity history, which downloads a file you can open to review it.
• Memories: go to Settings, then Personalization. Turning off Saved memories stops new memories from being saved; to remove existing ones, use Manage to delete single memories or select Delete all memories.
• Ads: under Settings, then Personalization, you can turn off Allow ads personalization. Microsoft says users who don't have a Microsoft 365 subscription might see ads in Copilot.
• Microsoft's pages we read do not say how long deleted chats remain in its systems after you delete them.

Government letters

These are letters, not findings or enforcement cases.

• Missouri: on July 9, 2025, the Missouri attorney general's office wrote to Microsoft about chatbot answers to a request to rank recent presidents on antisemitism; the letter says Copilot was among three chatbots that ranked President Trump last. Citing the office's authority under the Missouri Merchandising Practices Act, the letter says it is concerned that Microsoft's representations about its services to Missouri consumers are factually inaccurate, and it asked for voluntary answers and documents about how Copilot's responses are designed.
• 44 attorneys general: on August 25, 2025, a bipartisan group co-led by Illinois, North Carolina, South Carolina and Tennessee wrote to 13 companies, including Microsoft, asking them to adopt safeguards that keep their AI products from sexualizing children.
• 42 attorneys general: in a letter announced December 10, 2025 and led by Pennsylvania, the group wrote to 13 AI companies, including Microsoft, asking for safety testing, recall procedures, clear consumer warnings and other safeguards, and asked them to commit to changes by January 16, 2026.
• FTC: Microsoft was not among the seven companies that received the FTC's September 11, 2025 6(b) orders on AI chatbots that act as companions.

On the official sites we checked, we did not find a state or federal enforcement action against Microsoft over Copilot's consumer safety or privacy practices as of October 29, 2026.

Which laws may apply

No chatbot law names Copilot. Whether Copilot counts as a companion chatbot, an AI companion or a conversational AI service under a particular state law is a legal question. These laws set duties for chatbot operators:

• California SB 243, in effect since January 1, 2026: according to its author's office, companion chatbot operators must keep a protocol for responding to suicidal ideation, suicide and self-harm that refers users to crisis services, and must disclose that companion chatbots may not be suitable for some minors. A later law, SB 1119, signed September 10, 2026, removes SB 243's duties toward users an operator knows are minors, such as AI-status and break reminders and sexual-content measures, on January 1, 2027; SB 1119's own child safety duties begin July 1, 2027. The author's office says SB 243 lets families sue.
• New York's AI companion law (General Business Law Article 47): operators must have a protocol to detect and respond to expressions of suicidal ideation or self-harm and refer users to crisis services such as 988, and must tell users at the start of an interaction, and at least every three hours, that they are not talking to a human.
• Idaho's Conversational AI Safety Act (S 1297), signed March 31, 2026 and effective July 1, 2027: operators of public conversational AI services must disclose that the service is AI when people could be misled, adopt a suicide-response protocol with crisis referrals, and meet extra duties for minor account holders. The law excludes some tools, including features inside other software and tools marketed mainly for business use. The Attorney General enforces it.
• The two attorneys general letters described above each have their own record on this site, linked below. For comparison, see our guides to whether ChatGPT, Claude and Meta AI are safe.

We are not saying Microsoft is covered by, or out of compliance with, any of these laws. It may have practices it has not published, and whether a law applies to it is a legal question this guide does not decide.