Meta's privacy policy folds AI chats into its ad-data system

The design

Meta's AI characters and its Meta AI assistant run inside Facebook, Instagram, and Messenger, and conversations with them fall under the company's main Privacy Policy rather than a standalone document. As retrieved on 16 September 2026, the policy defines a data category called “Interactions with our features that are part of AI at Meta and related metadata,” explained as “content and messages you or others exchange with these features.” That category appears repeatedly across the policy's list of purposes for which Meta says it processes information, alongside long-standing categories like posts, comments, and device data.

What the evidence says

Under the heading “Personalizing Meta Products,” the policy lists AI-at-Meta interactions as one input used to personalise “features and content,” in the same bullet structure that separately lists “Personalize the ads people see.” A distinct section, “Developing and improving AI at Meta,” states the company uses the same interaction data to “develop and improve AI at Meta for Meta Products and for third parties.” Meta's Transparency Center adds, in its page on AI at Meta training data, that “information from people's interactions allows for continuous improvement of AI features” and that training datasets “may include information such as names or other details that people share publicly on our products or provide through their interactions with AI at Meta.”

What it asks of people

Talking to a Meta AI character or assistant means the content of that conversation sits inside the same general data-processing framework as the rest of a person's activity on Meta's apps, rather than a ring-fenced companion product. The policy does not separately ask for consent to use AI conversations for ad personalisation beyond the consent a person already gives by using Meta's apps under this policy, and it does not offer a companion-specific settings page distinct from Meta's general ad and AI controls.

Privacy and safeguards

The policy names “AI integrations” and a dedicated page on “how Meta uses information for generative AI models” as places for more detail, but the version reviewed here does not state a retention period specific to AI-persona conversations, nor does it describe a mechanism to remove a single past conversation from a training dataset once it has been used.

• Can a person separate their AI-character conversations from Meta's broader ad-personalisation system in practice?
• What does “for third parties” mean when the policy says AI at Meta is developed for Meta Products and third parties?
• Is there a way to see which past conversations were used to train a model, once training has happened?

Meta's disclosures show, in the company's own words, that AI-persona conversations are not treated as a separate privacy category from the rest of its products; they sit inside the same purposes list that includes advertising and model development, which is itself the notable finding for anyone assuming a chat with an AI character is walled off from Meta's ad business.