Big tech companies use user data to develop and improve AI models; however, many users do not know that by changing a few simple settings they can limit the use of their personal information in training these models.
Companies such as OpenAI, Google, Meta and LinkedIn rely on large volumes of data to develop AI technologies. This data may include users' conversations with chatbots, content posted on social networks, and some online interactions.
However, the way user information is used and the control options available are not the same across services. On some platforms, default settings may allow data to be used for AI training, and users must personally change the relevant privacy options to prevent this.
It is important to note that disabling the use of data for AI training does not necessarily mean deleting information that was previously collected.
Preventing the Use of Conversations in ChatGPT
OpenAI allows ChatGPT users to decide whether their conversations are used to improve AI models.
To disable this feature, users should go to Settings, select “Data Controls,” and then turn off “Improve the model for everyone.”
There is also the option to submit a data deletion request via OpenAI’s privacy portal.
However, there is an important point in the company’s policies. If a user provides feedback on ChatGPT’s performance by using the like or dislike buttons, the conversation related to that feedback may be used to improve models; even if the option to use conversations for training has been disabled.
Another solution is to use the “Temporary Chat” feature. Conversations conducted in this mode do not appear in the regular chat history and are not used to train models; although they remain subject to the company’s data retention policies.
Controlling Personal Information in Google Gemini
Google has also provided a way to manage the data used in the Gemini AI service. Users can go to their activity settings through the Gemini website or app and, by turning off “Keep Activity,” prevent future activities from being saved to the related history. It is also possible to delete previous conversations in the same section.
However, disabling this feature does not mean all information is immediately deleted.
According to stated policies, Google may retain some conversations that are not linked to an account for up to 72 hours for security review. Also, information related to conversations previously evaluated by human reviewers may be retained for up to three years.
How to Prevent Meta from Using Instagram and Facebook Posts?
Meta, the parent company of Facebook and Instagram, has a different mechanism for controlling the use of user data in AI training. Unlike ChatGPT and Gemini, which offer options to directly change settings, Meta provides a way to submit an objection request regarding the use of certain data through the Meta Privacy Center.
This request can limit the use of some data, including public Facebook and Instagram posts, comments, images, and interactions with Meta AI in future AI training; although the scope of this limitation depends on the company’s policies.
Meta has stated that it currently does not use users’ conversations with WhatsApp AI features to train its models.
Nevertheless, submitting an objection request does not mean completely preventing the processing of all information related to the user. For example, if someone else posts a public image and tags the user in it, Meta may still be able to process that image in accordance with its policies.
Disabling AI Training with LinkedIn Data
LinkedIn has also provided an option to control the use of user information in developing generative AI models. To access this feature, users should go to account settings, open the “Data Privacy” section, and disable “Data for Generative AI Improvement.”
This change can prevent the use of data covered by this setting in future training processes, but it does not delete information that was previously used to train models.
According to a reported statement, users’ private LinkedIn messages are also excluded from this type of use for AI training.
Is Changing Privacy Settings Enough?
According to Eitna citing NotebookCheck, reviewing and changing data-use settings on major platforms takes about 15 minutes. However, users should know where control options are located on each service and exactly what limitations each option creates.
The most important point is that preventing the use of information in future training is different from deleting data that was previously collected or used to train models.
Therefore, changing privacy settings can be an effective step to increase users’ control over personal information, but it does not necessarily mean completely removing their data footprints from AI systems.