Do AI companion apps train on your chats?

Usually, yes. Most AI companion apps train on your chats by default. They use conversation logs to improve their own models, or hand them to the third-party model provider the app runs on. More apps now offer an opt-out. It’s almost never switched on by default, and signup rarely mentions it.

Training itself isn’t the main problem. The problem is that you often can’t tell whether it’s happening to you. When Mozilla reviewed a batch of relationship chatbots, the pattern wasn’t apps loudly explaining their training practices. It was apps not explaining them at all.

Why do AI companion apps train on your chats in the first place?

AI companion apps train on your chats because those logs are the cheapest source of exactly the data they need. A general-purpose model can write an email. It doesn’t naturally know how people talk at 1am when they’re lonely, how they word an apology to a character, or which replies keep someone typing. Real chats teach that. Nothing else teaches it as well.

The second reason is blunter: retention. Logs show which replies end a session and which keep it going. Train on that and the product gets stickier. That’s a business goal, not a quality goal, and from the outside the two are easy to mix up.

Which parts of a conversation actually become training data?

The parts of a conversation that end up as training data cover more than most people imagine. It’s usually the whole exchange, not just your messages: your text, the character’s replies, edits and regenerations, any ratings or thumbs you gave, plus metadata like session length and time of day.

Two things catch people off guard. One: stuff you deleted from your screen may already be in a training set, because deleting something in the interface and deleting it from a data pipeline are two different operations. Two: anything you told the app to remember about you, the profile details that make a companion feel continuous, is structured personal data. It’s often stored separately from the chat log.

A 2025 incident cited in an August 2026 Congressional Research Service report shows what’s at stake. More than 300,000 Grok conversations were exposed. Some included names, passwords, and other personal details people had typed into what felt like a private box.

How do you opt out of AI companion training?

Opting out of AI companion training works on most major apps, but the switch is usually buried three or four screens deep. Go to Settings and look for a section called Data, Privacy, or Data Controls, then a toggle worded something like “improve the model for everyone” or “help improve our AI.” Turn it off. That’s the opt-out. Often there’s no confirmation and nothing in the product looks different.

Three practical points. The opt-out usually only covers the future, so it won’t pull old chats out of a model that already trained on them. Some apps lock the toggle behind a paid tier, one of the less obvious parts of what a subscription actually buys you. And if the companion app is a wrapper around someone else’s model, you may have to check both the app’s policy and the provider’s. For comparison, OpenAI documents its own training toggle publicly, which is more openness than most companion apps manage.

Can you tell whether a specific AI companion app trains on your chats?

To tell whether a specific AI companion app trains on your chats, read the privacy policy for two phrases: “improve our services” and “train our models.” The first one stretches to cover training without saying the word. The second is the honest version.

If the policy is vague, assume yes. Vagueness in this category isn’t an accident. Mozilla’s review of 11 relationship chatbots found most policies said very little about how conversation content feeds their AI, and only one app confirmed it doesn’t share or sell personal data. Ten of the eleven also failed Mozilla’s minimum security standards. That’s a separate warning, and just as important, about how carefully your chats are being stored.

Keep two ideas apart that often get blurred: a person reading your chats, and a model absorbing them. Companion conversations are generally private from other users but not from the company. Training is one specific thing the company might do with that access.

Does the law require AI companion apps to tell you?

The law is starting to make AI companion apps disclose more, but the coverage has holes. Data-protection rules like the GDPR and California’s privacy law give you the right to know what’s collected and to ask for deletion, and that includes chat logs. What they don’t do is force a plain-language notice the moment you start talking to a character.

Bills introduced in the 119th Congress would add companion-specific rules, including disclosure requirements and limits on processing user data for chatbot outputs. As of September 2026 they’re proposals, not law. So for now your real protection is the settings screen, not a statute.

What is still unknown about AI companion training?

What’s still unknown about AI companion training is most of the detail you’d need to make a fully informed choice. Companies rarely publish which conversations go into a training run, how long logs are kept before or after an opt-out, or whether human reviewers read samples. Independent audits of companion apps are rare, so outside evidence mostly comes from privacy reviews and breach disclosures, not verified internal practice.

Where this is heading isn’t settled either. Disclosure has improved since 2024, and several apps that had no training toggle now have one. But progress is uneven, and a policy can change in an update you never read. Don’t panic and don’t trust blindly. Check the setting, assume anything you type could stick around, and keep the truly sensitive stuff out of the box. At Vinfluencer we build AI conversational companion software and treat that last point as a design constraint, not a disclaimer.

Frequently asked questions

Does deleting my account delete my chats from the training data?
Deleting your account usually removes chats from your view and from active storage. A model that already trained on that text can’t unlearn it. Deletion requests under GDPR or CCPA cover stored records, not the weights inside a trained model. If you’re going to delete, do it early.

Do free AI companion apps train on chats more than paid ones?
Free tiers are more likely to train on your chats, since your data is the payment when money isn’t. Some apps sell the training opt-out as a paid feature. Paying doesn’t automatically stop training, though, so check the setting on any tier.

Can a human at the company read my companion conversations?
Sometimes, yes. Lots of privacy policies allow staff to review samples for safety, abuse detection, and quality checks. That access is usually limited and logged. It still exists on most platforms, and it’s separate from automated training.

Is anything I type genuinely off limits?
Nothing is off limits by default. So the practical rule is simple: don’t type passwords, financial details, medical record numbers, or identifying information about other people. A companion app doesn’t need any of that to hold a conversation.