The short answer: To opt out of AI training, change one setting at each company. For ChatGPT, turn off “Improve the model for everyone” under Data Controls. For Google, set Gemini Apps Activity to auto-delete at myactivity.google.com. For LinkedIn, turn off “Data for generative AI improvement.” For Adobe, turn off Content Analysis. Meta gives EU and UK users an objection form; most other users, including India, get no real opt-out. Public posts and old web content cannot be pulled back out of models already trained. The rest of this guide gives the exact path for each company below.
Jump to a company:
- Real opt-outs that actually work — OpenAI, Google, Meta, LinkedIn, Adobe, Apple, X, Spotify, Snapchat
- No opt-out available — and why
- For Indian users specifically
- What to prioritise
Why this matters
Every time you talk to an AI, you hand the company something valuable: a data point about how real people use language, ask questions, and describe problems. AI companies need enormous quantities of this to improve their models. You generate it for free. They profit from it indefinitely. That asymmetry is the deal you agree to when you sign up. Most people never read that part.
The good news is that several major companies added opt-outs after regulatory pressure. The bad news is that those opt-outs are buried, inconsistent, and often cover less than you’d expect. This guide covers every major company, what the toggle actually stops, and, just as importantly, what it doesn’t.
Real opt-outs that actually work
These settings exist, are relatively easy to find, and meaningfully reduce how your data feeds AI training. Change them.
OpenAI / ChatGPT
This is the one most people care about. By default, ChatGPT saves your conversations and uses them to improve OpenAI’s models. Here’s how to turn that off:
- Web: Settings (bottom-left) > Data controls > Toggle off “Improve the model for everyone”
- Mobile: Tap your profile icon > Settings > Data controls > Toggle off “Improve the model for everyone”
What this actually stops: your future conversations being included in OpenAI’s model training datasets. Conversations you’ve already had may have already been used.
What it does not stop: OpenAI retains your conversations for roughly 30 days regardless of this setting, then deletes them. (A 2025 court order in the New York Times lawsuit forced OpenAI to preserve deleted chats indefinitely. That order was lifted in late 2025, so the standard 30-day window is back as of mid-2026.) Human reviewers may still see flagged conversations for safety purposes. Metadata is still collected: when you used the service, how long sessions ran, and which features you used.
One note on plans: ChatGPT Team and Enterprise plans do not train on your data by default. This setting is only relevant on the Free or Plus tier. If your employer provides ChatGPT through an enterprise contract, you’re already protected on the training side.
Google is several products in one, and each has a slightly different situation.
Gemini Apps Activity: Go to myactivity.google.com > Gemini Apps Activity. From here you can turn this off entirely or set auto-delete to 3 months. Turning it off stops your Gemini conversations from being saved and used to improve Google’s AI models.
Google Workspace (work and school accounts): If you access Google through your employer or university, your data is not used for AI training by default. That’s managed at the administrator level, not your account.
Google Search: There is no opt-out here. Search queries have always been used to improve Google’s products, including AI features like AI Overviews. This has been baked into Google’s terms since the beginning.
YouTube: Your watch history feeds YouTube’s recommendation engine and its AI systems. You can pause this at myactivity.google.com > YouTube History. Note that pausing it doesn’t delete what’s already been collected — you need to do that separately.
Meta (Facebook, Instagram, WhatsApp)
Meta’s AI situation is the most complicated of any major company — and the most frustrating if you’re in India.
Meta AI (built into WhatsApp, Instagram, Facebook): EU and UK users have the right to object to their data being used for Meta AI training under GDPR. The setting is in Privacy settings on each platform under “Generative AI data settings.” You have to do this individually for each Meta app.
Indian users currently have no equivalent enforceable right. DPDPA gives you consent withdrawal rights in theory, but Meta has not built a functional opt-out mechanism for Indian users, and no enforcement action has been taken. The opt-out exists in Europe because regulators forced it there.
Advertising AI: This is separate from the generative AI training setting. Your browsing behaviour on third-party websites feeds Meta’s ad targeting AI. You can limit this by going to Facebook Settings > Privacy > Off-Facebook Activity and clearing or managing that data.
Image training from public posts: If your Facebook or Instagram account is public, your photos and posts fall under Meta’s terms of service, which include AI training rights. There is no opt-out for this. The only way to remove your content from future training is to make your account private or delete the posts.
LinkedIn rolled out a “Data for Generative AI Improvement” setting for US members in September 2024, switched on by default. It is one of the cleaner implementations, even though everyone was opted in first.
Go to Settings & Privacy > Data privacy > Data for generative AI improvement > Toggle off.
This stops LinkedIn from using your profile data, posts, connection activity, and interactions to train its generative AI features. The setting covers LinkedIn’s own models. It does not cover any data LinkedIn sells or licenses to third parties.
Worth doing. Takes about 30 seconds. Reporting at the time noted LinkedIn began using the data before it updated its terms of service, so opting out now stops future use, not past use.
Adobe
Adobe’s AI story is worth understanding because it caused genuine panic in 2023 — and the reality is more nuanced than the headlines suggested.
When Adobe updated its terms of service to include AI training rights, designers and photographers read it as Adobe claiming the right to train on anything in Creative Cloud. That’s not quite what happened. Adobe’s Firefly generative AI is trained on licensed Adobe Stock content and public domain material, not on Midjourney or Stable Diffusion-scraped datasets. Training on your personal Creative Cloud files requires a separate opt-in that Adobe eventually made clearer.
Regardless, here’s the opt-out: Adobe account settings > Privacy > Content Analysis > Toggle off.
This stops Adobe from analysing your creative files stored in Creative Cloud for AI training purposes. If you’re a photographer, graphic designer, or illustrator with years of work in Creative Cloud, this is the one to change.
Apple
Apple’s AI approach is architecturally different from most companies on this list, which is worth understanding.
Apple Intelligence runs primarily on-device for most tasks. When it does use cloud processing (Private Cloud Compute), Apple’s design explicitly prevents data from being retained or accessible to Apple employees. There’s no training data opt-out needed because the design claims data doesn’t persist.
Siri: This is the one to manage. Go to Settings > Siri > Siri & Dictation History > Delete Siri & Dictation History. Additionally, toggle off “Improve Siri and Dictation” if it appears in your settings (availability varies by region and iOS version).
Keyboard predictions: Settings > General > Keyboard > turn off “Predictive.” This stops your typing patterns from being used to improve the system keyboard model.
Apple’s privacy claims are more credible than most, but “designed not to retain” is still a design claim, not a legal guarantee. The Siri setting is still worth changing.
X (formerly Twitter)
X added a setting for Grok — xAI’s model, which is built into the X platform — in 2023.
Go to Settings > Privacy and Safety > Grok > Toggle off “Allow your posts as well as your interactions, inputs, and results with Grok to be used for training and fine-tuning.”
What this covers: your direct interactions with Grok inside the X platform.
What this does not cover: your public posts on X. Public tweets are accessible under X’s terms and have been used for training by various AI companies. OpenAI even had a data licensing deal with X before it ended. Turning off the Grok setting doesn’t change any of this. If you post publicly on X, that content is effectively in the public training pool.
Spotify
Spotify uses your listening history to improve its AI recommendation and discovery models.
Go to Settings (on mobile: tap the gear icon) > Privacy settings. Toggle off “Using your data to improve Spotify” and “Tailored ads.” These are separate toggles covering different purposes, and you want both off.
This is one of the less consequential settings on this list for personal-information exposure. But Spotify knows more about your mood, routines, and emotional state than most apps. Listening patterns at 2 AM tell a different story than what you’d willingly share.
Snapchat
Snapchat’s My AI feature is the chatbot built directly into the app. It drew significant regulatory attention in 2023. The UK’s ICO investigated whether Snapchat had run an adequate data protection impact assessment before launching it to users, including children.
The opt-out: Settings > Privacy controls > Snapchat AI Data toggle.
Worth noting: My AI was pushed to the top of every user’s chat list as a default in 2023, and users had to actively dismiss it. Regulators and privacy advocates flagged the opt-in framing as problematic. The opt-out exists, but its existence was the result of external pressure, not proactive design.
No opt-out available — and why
Some things you cannot opt out of, regardless of how many settings you change. It’s worth being honest about this.
Your public social media posts. If your account is public on any platform — Facebook, Instagram, X, Reddit, TikTok — your posts are legally accessible under most interpretations of copyright and fair use law. Meta, X, Reddit, and others have licensing agreements with AI companies covering public post data. Making your accounts private is the only meaningful protection.
Web content you’ve published. If you have a blog, a website, forum posts, or public comments, they are almost certainly in AI training datasets. You can implement robots.txt with directives like User-agent: GPTBot and Disallow: / to block future crawling by OpenAI’s crawler and some others. But many crawlers ignore robots.txt, enforcement is non-existent, and if your content was crawled before you added the directive, it’s already in the dataset.
Phone number and contact data. Once your number is in Truecaller’s database — either because you registered or because someone in your contacts did — it’s being used for Truecaller’s AI-based spam detection models. Truecaller does have a deregistration process at truecaller.com/unlisting, but contact data shared by others is not something you can unilaterally delete.
For Indian users specifically
The legal picture in India deserves a direct conversation.
The Digital Personal Data Protection Act (DPDPA) passed in 2023 and gives you the right to withdraw consent under Section 11. In theory, you can invoke this to ask AI companies to stop processing your personal data. In practice, no Indian enforcement action has been taken against a foreign AI company as of mid-2026. The Data Protection Board of India was established in law in November 2025. But as of mid-2026 it is not yet fully operational: leadership appointments are still pending, and most substantive enforcement provisions take effect only from 2027. There is also a structural gap. The DPDPA excludes personal data you have made publicly available, so posts on a public account may fall outside its protection entirely. Companies know all of this.
GDPR gives EU residents substantially stronger rights, including the right to object to data processing under Article 21. If you travel to EU countries, work with EU-based companies, or use services with EU operations, you may have access to these rights depending on the service’s legal structure. Many companies have built GDPR-compliant opt-outs that only activate based on detected location, not account settings.
The practical conclusion for Indian users: treat opt-outs as partial measures and genuine risk reduction rather than legal guarantees. The most effective protection is limiting what you share in the first place — not just how it gets used afterward.
What to prioritise
If you do nothing else after reading this, do these four things:
- ChatGPT: Turn off “Improve the model for everyone” in Data controls.
- LinkedIn: Turn off “Data for generative AI improvement” in Data privacy settings.
- Google: Set Gemini Apps Activity to auto-delete at 3 months at myactivity.google.com.
- Adobe: Turn off Content Analysis in Adobe account Privacy settings if you store creative work in Creative Cloud.
Meta is the most important one if you’re in the EU. It is the most frustrating if you’re in India, where the opt-out mechanism doesn’t fully exist yet in the same enforceable form.
The honest version of all of this: opt-outs reduce your training-data exposure at the margins. They don’t eliminate it. The companies that built them did so because regulators forced them to. They are designed to be just functional enough to satisfy compliance requirements. The real question isn’t only “how do I opt out of AI training.” It’s “how much personal information am I generating, and where is it going?”
AI training is only one route your data travels. The same details also sit on people-search brokers that resell your name, address, and phone. Broker pages like Spokeo and BeenVerified show what they hold and how to opt out.
What you can do right now
Want to see what data brokers expose about you? See what a scan finds → shows an example report: the kinds of sites that list your name, phone, and address. Once KnytX Shield launches it will find and remove these automatically. Join the waitlist to get in early.