News chronological

Approaches vary among services: some offer temporary chats that are not used for memory or model training; some services only train on user data if the user gives explicit permission, while others require the user to opt out to prevent their data from being used for training. Additionally, many services allow users to view stored data and delete specific memories, sometimes through simple requests to the chatbot.

Miranda Bogen suggests that the more a system knows about a user, the easier it becomes for the system to make escalating requests for private details in a manner that feels natural. She notes that without robust privacy protections, the incentive for companies to monetize that knowledge will be difficult to resist.

Google stated that it will use photos and other materials that people upload through Search to train its AI systems by default, although users do have the ability to opt out.

OpenAI announced an optional Computer History feature that allows ChatGPT to maintain a record of the various apps and websites a person uses.

Meta states it can use interactions with Meta AI to personalize content and ads across its services, including certain interactions through its smart glasses. However, Meta specifies that it does not use conversations regarding sensitive topics—specifically health, politics, and religion—to personalize ads. Meta has also introduced an "Incognito Chat" mode for temporary, private conversations where Meta's servers see the query only to provide an answer and do not store the data.

Apple offers a private option by handling requests on-device whenever possible. For more demanding tasks, Apple utilizes a system called Private Cloud Compute, which is designed to use data only to fulfill the specific request and ensures the content does not become accessible to Apple. However, this privacy approach does not apply if a user chooses to send a request to a third-party service like ChatGPT.