Give it a shot: Sign into ChatGPT or Meta AI and enter, "roast me using my conversation record." If it has sufficient knowledge of you to deliver a sharp critique, that could indicate it's time to enhance your artificial intelligence privacy measures.
Many artificial intelligence firms maintain records of all your interactions with their chatbots. The content you provide—your speech, images, actions, and thoughts—contributes to enhancing their AI systems. They leverage this data to tailor bot replies, encouraging extended engagement from you. Moreover, these conversations also support their additional ventures; for instance, this week, Meta has begun utilizing information shared with its bot to deliver targeted advertisements on Instagram and Facebook.
Spend 15 minutes with me, and I'll show you how to safeguard your information. I examined the settings of the top AI platforms and selected the least secure default privacy options that require adjustment. The links and images provided will guide you directly to the necessary changes for OpenAI’s ChatGPT, Anthropic’s Claude, Microsoft’s Copilot, Google’s Gemini, and Meta AI.
What makes it important whether an AI firm holds information about you?
- You may prefer not having advertisements track you according to your more private discussions.
- When artificial intelligence firms use your data to develop their tools, we remain uncertain about the effectiveness of their methods in removing names, facial features, addresses, and other sensitive details — or how frequently such information resurfaces in responses generated by AI systems.
- Artificial intelligence assistants, designed to perform actions for you, may encounter pitfalls or become unpredictable, potentially causing significant harm through access to your credit card details, passwords, or private data.
- Legal professionals and governmental authorities may ask for your messages as proof or clues.
Bots are currently creating what are known as memory files from your conversations to customize their replies. However, they make mistakes—my ChatGPT memory stated I had a spinning class. They might recall details you wish they could erase, such as confidential health issues. Additionally, bots acting as friends who know you well have the potential to influence individuals and may contribute to psychological disorders.
I developed the following guidelines after discussing with three experts in AI privacy: Jennifer King from the Stanford Institute for Human-Centered Artificial Intelligence, Jacob Hoffman-Andrews of the Electronic Frontier Foundation, and Ben Winters representing the Consumer Federation of America. Every bot has some form of privacy concerns. A general guideline: Avoid having a chatbot capture information that you would not be comfortable sharing on social media.
(Technology firms often change locations, so if you notice one has moved, please send me an email.)
Ensuring your privacy when using AI tools involves certain compromises. The main one being that the more limited a chatbot’s knowledge of you is, the less tailored its replies may become. However, I believe these bots aren’t companions — and I have suggestions for maximizing AI benefits while avoiding monitoring.
- - -
If you accomplish just one task If you focus on a single action If you carry out only one step If you perform one specific act If you take care of one matter If you complete one activity If you handle one issue If you undertake one endeavor If you execute one action If you manage one task
One action can safeguard you without altering any other configurations: opt for "temporary chats."
They function similarly to "incognito" mode in a web browser. Temporary conversations prevent the discussion from being stored in your history, stopping it from contributing to a bot's recollection of you or being utilized to improve artificial intelligence later on.
Utilize ephemeral conversations for any matters that could be considered delicate, such as those related to health, finance, and personal relationships. Drawback: Your discussion will not be stored, making it impossible to revisit or carry forward at a later time.
This is what the buttons appear as:
Regrettably, Meta AI does not offer a temporary conversation feature, however, you have the option to sign out from the application or site and pose inquiries in a manner that will not be recorded under your profile. This applies similarly to the individual edition of Copilot.
Let's now adjust some configurations that will ensure the safety of all your conversations.
- - -
ChatGPT
Out-of-the-box, the frequently used chatbot retains copies of every conversation you have and creates an individualized set of recollections regarding you. The developer of ChatGPT, OpenAI, additionally allows itself to utilize your input for training subsequent iterations of ChatGPT. It seems probable that advertisements tailored to your discussions may not be too distant, although the company mentions it will "approach this thoughtfully." (The Washington Post maintains a content collaboration with OpenAI.)
What you need to do right now:
- Remove previous conversations. Navigate to Settings → Data Controls and select "Delete All." Regrettably, you can only do this retrospectively, and only if you recall. ChatGPT does not provide an option for automatic deletion of chats at regular intervals.
- Disable memory. Navigate to Settings → Personalization and switch off "Reference saved memories." A more private method of personalizing ChatGPT involves going to the same section and entering some "Custom instructions" detailing how you prefer the bot to reply. My version states: "request clarification before responding if a query is unclear, and provide citations when applicable."
- Prevent ChatGPT from using your data for training, as you will not benefit from it and it won't affect the accuracy of your answers. Navigate to Settings → Data Controls and switch "Improve the Model" to off.
- - -
Claude
Anbot from Anthropic, favored by developers and advanced users, maintains a record of every conversation you have, but it seeks approval prior to utilizing your information for training its artificial intelligence system.
What you need to do right now:
- Remove previous conversations. Go to the end of your recent chat list and click All chats. From there, you can choose multiple chats simultaneously and remove them. Claude does not provide an option for automatic deletion of chats according to a schedule, nor does it allow you to prevent keeping any conversation records.
- Prevent Claude from learning from your interactions. Navigate to Settings → Privacy and switch "Help improve Claude" to off.
- If you possess a premium subscription, disable the memory feature. Navigate to Settings → Capabilities and switch "Generate memory from chat history" to off.
- - -
Gemini
By default, Google stores your Gemini chat history for 18 months and uses this data to train its artificial intelligence models. However, it claims that it hasn't used these discussions for targeted advertising — at least not yet.
Gemini has also been incorporated directly into Gmail and Docs, leading to various types of worries.
What action you should take right now:
- Automatically delete previous chats. This feature is more effectively handled by Gemini compared to others. Go to Settings & Help → Activity. The optimal option is to turn "Keep activity" off, preventing Google from storing data about you and stopping it from utilizing your conversations to enhance its AI. If you want to retain certain old messages for future use, configure them to auto-delete after a specific period, like three months. However, be aware: There’s no method to prevent Google from using your text for training its AI if you maintain it in your chat history.
- Disable its memory. Navigate to Settings & help → Personal context and switch off "Your previous conversations with Gemini." An improved method to customize Gemini is to visit Settings & help → Instructions for Gemini and specify how you would like the bot to adjust its responses.
- Choose whether to disable Gemini integrated within Gmail and Docs. According to Google, it does not utilize the content of your emails or documents for training Gemini. Additionally, when you engage with Gemini in these apps, Google does not retain any information related to that interaction, which means there is no adjustable option available. However, if you're uncomfortable with Gemini accessing your private data, you have the ability to turn it off. Navigate to Gmail → Settings → General, then scroll down to Google Workspace smart features, click on Manage Workspace smart feature settings, and toggle it off. This applies to Gmail, Docs, Calendar, and Drive.
- - -
Meta AI
Meta's internally developed chatbot is difficult to escape: it functions as an independent application and website, and is also integrated within Instagram, Facebook, and WhatsApp. Additionally, Meta offers several methods for controlling how much information it retains or how the company utilizes your data. One thing to note: there isn't an option available to prevent it from using the content of your messages for targeted advertising purposes.
What you need to do right now:
- Remove previous conversations. On the Meta.ai website or application, navigate to Settings → Data and Privacy → Manage Your Information → Delete All Chats and Media. Meta does not provide an automatic method for regularly deleting chats.
- Erase your public content. Certain individuals have accidentally shared their Meta AI conversations and AI-created images and videos publicly. To ensure you haven't done the same, navigate to Settings → Data and Privacy → Manage Your Information → Remove All Public Posts.
- Check and remove your history. There’s nothing you can do to prevent Meta AI from storing information about you. However, you have the option to erase specific records. Navigate to Settings → History and choose the entries you wish to eliminate.
- - -
Microsoft
This can become somewhat unclear: There are various Copilots designed for individual use compared to those used by company staff. Additionally, they provide distinct privacy settings.
Additionally, it should be mentioned that the individual edition of Copilot displays advertisements according to the content of your conversations—unless you disable this feature.
What you need to do right now:
For individual users, covering those with either free Copilot accounts or paid Microsoft 365 memberships:
- Remove your conversation records. Navigate to your Microsoft account → Privacy → Copilot → Copilot applications → Erase all activity logs. There isn’t an option for automatic deletion following a specific duration. As you're there, also remove your usage of Copilot within other Microsoft programs from the same section.
- Ask Microsoft to cease using your conversations for targeted advertisements. Navigate to your Microsoft account → Privacy → Personalized Ads & Offers and set it to off.
- Disable personalization and data storage. Visit the Copilot site, click on your name located at the lower-left side, then select your name once more from the appearing menu. Navigate to Privacy settings, scroll down to Personalization and Memory, and turn it off. Additionally, tap on Delete Memory to remove previously saved information.
- Disable model training for text and speech. Visit the Copilot website, click on your profile at the lower-left corner, then select your name from the pop-up menu. Next, navigate to Privacy settings and set model training to disabled.
Corporate users have stricter privacy settings by default when it comes to training models with their data and utilizing it for advertising purposes. However, you should avoid entering anything into an enterprise version of Copilot that you wouldn't be comfortable showing your manager—since all information is owned by your organization.
Here's what you need to modify:
- Clear your conversation records. Navigate to your Microsoft account → Settings & Privacy → Copilot activity history and select Delete history. This will remove it from your display, although your IT admin may still have access to a backup.
- Disable personalization features. Access the M365 Copilot website and locate the three dots at the top-right corner, then select Settings → Personalization. Toggle off Saved memories and chat history. An improved method to customize Copilot’s behavior is through the same menu under Custom instructions, where you can specify how you'd like the AI to respond to your queries.
Posting Komentar