Artificial intelligence has become a core part of how the world’s most popular apps function. Email filtering in Google’s Gmail, content recommendations on Instagram, and music discovery on Spotify all rely on AI systems working behind the scenes. For users concerned about how their personal data is used to train these models, the reality is sobering: in most cases, AI cannot be fully switched off. What users can do, however, is place limits on what data is collected and how it is used.
Across major platforms, AI is deeply embedded in core services. In Google products such as Gmail, Drive, Docs and YouTube, AI handles spam filtering, smart replies, video recommendations, text autocompletion and file organisation. These systems cannot be disabled entirely, but users can reduce data collection by pausing Web & App Activity, Location History and YouTube History in the “Data & Privacy” section of their Google account. Disabling ad personalisation and deleting older activity records can further reduce long-term data retention.
Social networks owned by Meta, including Facebook and Instagram, use AI to curate feeds, personalise advertising, detect harmful content and train generative AI tools. While these functions remain active, users in the European Union have the right to object to the use of their data for AI training. This can be done through account settings, where users can limit ad personalisation and submit an objection form related to AI model training. Even with these steps, AI-driven content ranking and moderation continue to operate.
On WhatsApp, Meta’s AI assistant appears as a blue circle, offering features such as search, translation and content generation. The chatbot does not analyse user data unless it is directly interacted with, and private messages remain protected by end-to-end encryption, meaning they are inaccessible to Meta AI.
Apple takes a different approach. On iOS and macOS devices, AI features used in Apple’s Mail, Photos and Siri largely operate on-device. Facial recognition, photo classification, keyboard predictions and app suggestions are processed locally, minimising data sent to Apple’s servers. For tasks that require cloud processing, Apple says data is encrypted, not directly tied to user identities and deleted after use. The company also says it trains many of its AI models using synthetic data rather than real user content.
On X, owned by Elon Musk, AI is used for moderation, recommendations and to train the Grok generative model. Users can opt out of having their data used for AI training, but public posts are still analysed automatically to ensure the platform functions as intended.
Spotify relies heavily on AI to generate personalised playlists and music recommendations, features that cannot be disabled without undermining the service itself. Users can, however, turn off personalised ads on the free tier and exercise data access or deletion rights under EU privacy rules.
While switching off AI entirely is rarely an option, regulators and privacy experts note that understanding platform settings and exercising available controls can significantly limit how personal data is used. As AI becomes inseparable from digital services, managing its boundaries is increasingly a matter of user awareness rather than total avoidance.
