Meta expands Instagram privacy controls to curb off-platform data use for AI and ads
A new "Activity from other businesses" setting allows users to limit how external data influences their feed and AI responses, though in-app tracking and model training remain largely unchanged.
Meta has begun rolling out a broader privacy control in Instagram titled "Activity from other businesses," marking a significant shift in how the platform manages external data for advertising and artificial intelligence. Starting in July 2026, this new setting replaces the previous "Your activity off Meta technologies" and expands the older "Activity information from ad partners" control. The update allows users to prevent third-party businesses from using data they have already sent to Meta for the personalisation of ads, Feed content, and AI responses.
While the new control offers greater granularity, it does not act as a master switch to halt all data collection. Meta continues to gather information from user activity within Instagram, including searches, likes, follows, posts, and clicks, as well as device, network, and location signals. Consequently, ads can still reflect the accounts users follow, the content they engage with, and customer lists uploaded by advertisers. The setting specifically ensures that activity occurring off Meta’s platforms does not inform the user’s ad experience, rather than shutting down all ad personalisation entirely.
In the European Region, users have additional options through the Accounts Center, where they can select "less-personalised ads" under Ad preferences. This option uses fewer signals than standard personalised ads, although Meta notes that it can still utilise details such as age, location, device information, and ad interactions during the current session. For users on iPhones, Apple’s "Ask App Not to Track" control can block cross-app tracking for advertising, but it does not stop Instagram from recording activity inside its own app. Android users can similarly manage permissions for location, contacts, and other data via system settings.
Meta has also integrated interactions with its AI features into its personalisation signals, a practice that began on 16 December 2025 in most regions. A conversation with Meta AI about a specific topic, for instance, could contribute to later recommendations for related posts or products. To manage this, Instagram offers two new cleanup commands: `/reset-ai` to delete a specific AI’s copy of the conversation, and `/reset-all-ais` to reset every AI chat in the app. These commands clear saved context but do not serve as opt-outs from advertising, ordinary data collection, or model training.
Model training remains a distinct process from personalisation. Meta uses public posts, comments from adult accounts, and interactions with its AI features to improve generative AI models in regions where it relies on legitimate interests. Users in the European Union and the United Kingdom can object to this use through Meta’s Privacy Center, a right that is not available as a universal Instagram setting. The company advises users to avoid sharing sensitive details, such as names, addresses, or financial information, in Meta AI chats, as these interactions may be used for model improvement.
The rollout is proceeding by country, meaning some accounts may still display the older "Activity information from ad partners" setting under Ad preferences. As Meta continues to refine its data practices, the distinction between in-app activity, off-platform data, and AI model training remains a critical area for users seeking to manage their digital footprint.
