Change log
What changed, when, and the official source.
| Date | Tool | Change |
|---|---|---|
| Date2026-11-19 | ToolGemini | ChangeCorrected the Gemini training default. Google's help page says "Chat content including text & attachments" is on by default if you are over 18, so you need to uncheck it to opt out. "Audio and Gemini Live videos & screenshares" is off by default. History is also on by default if you are 18 or over. To turn chat-content training off: Settings → History → uncheck "Chat content including text & uploads" → Turn off. Google uses "attachments" in the help text and "uploads" in the steps. Two-month retention when History is off is unchanged. Source |
| Date2026-11-16 | ToolGemini | ChangeGoogle is splitting Gemini chat history from generative AI model training. New settings let you keep your Gemini chat history while opting out of generative AI model training. If you turn the new History setting off, future chats are saved for around two months. Temporary chats are kept for around two months too. Reviewed chats can be kept for up to 3 years. Google will start rolling these changes out globally in the Gemini app in the coming weeks. Blog: https://blog.google/innovation-and-ai/products/gemini-app/updated-controls/ Privacy Hub: https://support.google.com/gemini/answer/13594961 Source |
| Date2026-11-12 | ToolSmule | ChangeNew guide for Smule’s AI-training and partner-licensing opt-outs. The updated terms and privacy policy take effect December 4, 2026, and set the minimum age at 18. Source |
| Date2026-11-12 | ToolGemini | ChangeGoogle's Connected Apps page lists Contacts as “Contacts, which includes Google Contacts and contact info from your signed-in device.” That line no longer adds “US only, English only, and requires an AI Ultra subscription.” Leftover Gemini Apps Activity can still be used for product improvement, including training generative AI models, and deleting it prevents future use unless trained reviewers have already reviewed it. Source |
| Date2026-11-12 | ToolGmail, Docs & Photos | ChangeSame Connected Apps update: Contacts is listed without the former “US only, English only, and requires an AI Ultra subscription” limit, and Google says leftover Gemini Apps Activity can still be used for training after you disconnect an app unless you delete that activity (except data already reviewed by trained reviewers). Source |
| Date2026-11-12 | ToolChatGPT | ChangeOpenAI's deleting-and-archiving article says that if a chat used connected-app data, archiving keeps that in-chat copy, deleting the chat removes it from the account but not the original app data, and disconnecting an app does not delete past conversations that already used app data. If ChatGPT ends a conversation over a serious concern about violence or harm, that conversation is no longer in visible chat history. Source |
| Date2026-11-12 | ToolChatGPT | ChangeOpenAI says it is testing a multi-image visual ad format during image generation. The ad is clearly labeled and separate from the image you asked ChatGPT to create. Source |
| Date2026-11-11 | ToolClaude | ChangeNoted that Claude Fable 5, 5.1, and 5.2 are Covered Models. Where zero data retention would otherwise apply, Anthropic keeps prompts and outputs at least 30 days for safety monitoring and does not offer zero data retention unless it expressly authorizes it. Consumer Free, Pro, Max, and Ultra plans, including Claude Code, are unaffected. Source |
| Date2026-11-11 | ToolGitHub Copilot | ChangeNoted that Claude Fable 5.2 is in GitHub Copilot for Pro+, Max, Business, and Enterprise (gradual rollout). Using Fable 5, 5.1, or 5.2 means Anthropic retains prompts and outputs for safety classifiers, outside GitHub's usual zero-data-retention agreement unless an eligible enterprise is approved. Business and Enterprise admins must enable the Fable 5.2 policy, which is off by default. Source |
| Date2026-11-11 | ToolCursor | ChangeNoted that Claude Fable 5.2 requires Anthropic to retain inputs and outputs even with Privacy Mode on. Cursor says that data is not used to train or improve Anthropic's models or products, and Privacy Mode or Enterprise accounts must approve the retention policy before use. Source |
| Date2026-11-06 | ToolGranola | ChangeNew guide covering the model-training opt-out and Default link sharing (Anyone with the link vs Private / Only my company). Source |
| Date2026-11-05 | ToolMicrosoft Copilot | ChangeMicrosoft now says people who used the older Copilot app must sign in to the updated app by December 31, 2026 to keep eligible chats and content; content that isn't migrated will be permanently deleted after that date. Source |
| Date2026-11-05 | ToolChatGPT | ChangeAdded OpenAI's notes that archiving or locking a chat doesn't change training, deleted chats can't be restored, Temporary Chats don't show ads, ads aren't shown to accounts identified as under 18, and a free no-ads option with limits exists in some regions. Source |
| Date2026-11-05 | ToolAll | ChangeWeekly check: 22 pages re-checked against official sources; 1 vendor change (Microsoft Copilot), 1 clarification (ChatGPT). |
| Date2026-11-05 | ToolAll | ChangeNine older guides updated from the Nov 1 freshness pass (Claude, Gemini, Perplexity, Lovable, Meta AI, Muse, Grok, GitHub Copilot, Cursor). Sources were re-read Nov 5 with no further change. GitHub Copilot and Cursor had no factual change. |
| Date2026-11-01 | ToolGemini | ChangeGoogle's Gemini Privacy Hub (updated Oct 26, 2026) says chats and memory you import from other AI apps are saved to your Activity and can be used to train its models. Source |
| Date2026-11-01 | ToolPerplexity | ChangeAdded Perplexity's official account-deletion steps and Incognito mode. Source |
| Date2026-11-01 | ToolMuse | ChangeFixed the download-data label to "Download your Muse data" (Settings → Data controls). |
| Date2026-11-01 | ToolLovable | ChangeNoted that trained team members may review content to check model quality. Source |
| Date2026-11-01 | ToolMeta AI | ChangeNoted that, depending on your region, interactions with AI products can be used to personalize content and ads. |
| Date2026-11-01 | ToolNew guides | ChangeAdded guides for LinkedIn, Grammarly, Canva, Pinterest, Google Search (Save Media), Siri & Apple Intelligence, Gmail, Docs & Photos, WhatsApp, Alexa, Instagram & Facebook, and Twitch. Also added a page on Reddit, TikTok, Spotify and CapCut, where no AI-training opt-out was found on the official pages checked. |
| Date2026-11-01 | ToolMicrosoft Copilot | ChangeMicrosoft's documentation for the new Copilot app (August 18, 2026) says prompts, responses and file contents aren't used to train foundation models. The training switches on our page now apply only to the older app. Source |
| Date2026-11-01 | ToolChatGPT | ChangeOpenAI's data-controls article says Trusted Access sessions, such as Daybreak Blue, may still be used to improve models when “Improve the model for everyone” is off, and that voice mode has a separate training control. Related help pages now describe temporary chats as a copy kept up to 30 days for safety purposes, a de-identified-chat exception on deletion, and ad controls for the Free and Go plans. Source |
| Date2026-10-13 | ToolAll | ChangeInitial pages drafted from official vendor documentation. Each page lists its sources and the date checked. |
| Date2026-09-09 | ToolLovable | ChangeLovable's docs say Free and Pro customer data may be used for model training as of this date unless you opt out. Source |
| Date2026-08-18 | ToolMicrosoft Copilot | ChangeMicrosoft notes an updated Copilot app with a separate privacy-controls article. Our page is being verified against it. Source |
| Date2026-04-24 | ToolGitHub Copilot | ChangeGitHub's docs say Free, Pro, Pro+ and Max interactions may be used for AI training starting this date unless disabled. Source |