AI Memory: A New Feature That’s Becoming More Widely Available

AI Memory: A New Feature That’s Becoming More Widely Available – How It Works and How to Enable It on Different Platforms

Not long ago, I complained in one of my posts that AI memory is basically “2-bit” (in my own words 😀), which can be really annoying, especially when you’re working on a longer process or project. The good news is that this very useful AI memory feature is now starting to spread across different platforms.

It’s not available for every user right away (although in the paid version I’ve had it for a while now), as it’s being rolled out gradually depending on region and account type. This applies to both the Microsoft Copilot Memory feature and ChatGPT Memory, so while you might not see it yet, don’t worry – it should become available soon for most paid and free users as well. (Of course, if you use an AI service without registering for an account, this feature will not be available due to privacy reasons.)

In this post, we’ll look at how the memory feature works in ChatGPT, Microsoft Copilot, GitHub Copilot, and Claude, and how you can enable it:

1. ChatGPT – The ChatGPT memory feature allows the system to remember important information about you, such as your name, favorite topics, previous questions, or projects. This means ChatGPT can provide more personalized responses and follow what you’re interested in over the long term. How to enable it? On desktop: click on your profile picture → Settings. On mobile: profile → Settings. Go to the “Personalization” menu. Turn on the Memory feature and confirm that you want to use it. After that, ChatGPT will gradually start remembering information, which you can view or delete anytime. According to OpenAI’s official statement, they do not explicitly collect PII (Personally Identifiable Information). However, if you type in sensitive data such as names, addresses, phone numbers, or passwords, the system will process it during the conversation. The model improves using aggregated, anonymized data, not tied to individual users. Tip: Never enter passwords, credit card information, health documents, or company secrets! Another important thing: when ChatGPT Memory is enabled, saved information is visible in your settings. If someone gains access to your account, they will also be able to see your saved memories and preferences. This is why two-factor authentication (2FA) is strongly recommended for your ChatGPT/OpenAI account. 🚷

2. Microsoft Copilot “Memory” (personalization) feature is also starting to roll out and enables MS Copilot to remember how you work, what topics interest you, and what projects you are working on, to provide a more personalized experience over time. It works very similarly to ChatGPT Memory: it can remember information about you, like your name, favorite topics, or important parts of previous conversations, but only if you allow it. This feature is also appearing gradually for users. The goal of Microsoft Copilot Memory is to understand how you work, which apps and projects you use, and adjust its suggestions accordingly. How to enable it? Open Copilot settings on the web or in the app. Go to the Personalization or Memory menu. Turn on the memory feature. If you don’t see this option, the feature is not yet available for your region or account type.

3. GitHub Copilot – Context, not classic memory. In the case of GitHub Copilot, there is no traditional memory feature, but it can take the project context into account, making suggestions more relevant. With Workspace Indexing, Copilot can understand all files in the project (I personally use Visual Studio Code for this). Copilot for Business can allow the system to use patterns from private repos, but these are not added to the global model. This is more about context handling than true long-term memory.

4. Claude 4, Sonnet 4, Opus 4 models currently do not have a memory toggle feature, and there is no official news about this coming yet (though enterprise users are expecting it, so it’s possible that it might arrive there first). However, these models already “remember” better by default (and they are more expensive), as they are mainly used for development purposes. Promises have been made that the memory capacity will continue to grow, making the models better even without a dedicated memory feature. In my opinion, for development, Claude models are already the best, though for everyday problem-solving ChatGPT models are still the recommended choice.

Summary: AI memory features are still quite new and are being rolled out gradually. If you don’t see them in your settings yet, don’t worry – they will likely appear for you soon. In the future, these features will allow various AI tools to provide even more personalized and more time efficient assistance.