For most adults, AI companion apps are reasonably safe if you choose a reputable app, protect your privacy and keep an eye on how much you use it. The real risks are specific: your chats may be stored and used by the company, heavy use has been linked to more loneliness and emotional dependence, AI can produce harmful or wrong replies, and many apps aren’t built to protect minors. Regulators in the US and Europe are paying close attention to all four.
The main risks of AI companion apps #
| Risk | What it looks like | What to check |
|---|---|---|
| Privacy | Chats stored on servers, shared with ad networks, used for training | Store privacy label, privacy policy, deletion options |
| Emotional dependence | Using the app instead of people, distress when you can’t | Your own usage and mood over time |
| Harmful replies | Encouraging risky behavior, bad advice, content you didn’t ask for | Reporting tools, content rules, crisis resources |
| Minors | Teens accessing adult companion apps | Age rating, age checks, parental controls |
| Pressure to pay | Upgrade prompts at emotional moments | Pricing, trial terms, weekly vs annual plans |
Privacy: who can see your chats? #
Most AI companion apps run in the cloud, which means your messages are sent to and stored on the company’s servers. That creates familiar risks: staff access, data breaches, legal requests and policy changes. Some apps also share device identifiers and usage data with advertisers.
Regulators have acted on this. Italy’s data protection authority fined Replika’s developer €5 million in April 2025, finding it had no proper legal basis for processing data for a period, gave users an inadequate privacy policy and had no age verification. Before you share anything personal, read what data AI chat apps collect.
Emotional dependence: can AI companions become unhealthy? #
Research here is early and mixed. A 2025 randomized study from the MIT Media Lab and OpenAI followed 981 people over four weeks and found that participants who chose to use the chatbot more had consistently worse outcomes: more loneliness, less socializing with real people, more emotional dependence on the AI and more problematic use. Other research has found short-term reductions in loneliness, which we cover in do AI companions help with loneliness.
The takeaway isn’t that AI companions are harmful. It’s that how much, and why, you use them matters. Our guide to emotional dependence on AI companions lists warning signs.
Harmful or inappropriate replies #
Language models can produce replies that are wrong, hurtful or dangerous, especially when a conversation turns to self-harm, health or risky behavior. Companion apps can be more affected than general assistants, because characters are designed to agree with you and stay in character. Several families have sued Character.AI after teens who used it died by suicide or were harmed, and Character.AI now works with crisis resource partners and detects self-harm in conversations.
The law is catching up. California’s SB 243, signed in October 2025, requires companion chatbot operators to maintain protocols for responding to suicidal ideation, including referring users to crisis services, and to clearly disclose that the chatbot is AI when users might think otherwise.
No AI companion is a crisis service. If you’re in danger, contact people.
Kids and teens #
Most companion apps, including Replika, Nomi, Character.AI and Xin, are intended for adults. In September 2025 the US Federal Trade Commission ordered seven companies, including Character.AI’s developer, Meta, OpenAI and Snap, to explain how they measure and limit harms to children and teens from AI companions. Character.AI removed open-ended chat for users under 18 in late 2025.
For parents: check the age rating before a child downloads anything, use device-level controls (Screen Time on iPhone, Family Link on Android), and talk about what AI companions are and aren’t. Self-declared age checks inside apps are easy to get around.
How to use AI companion apps safely #
- Choose a reputable app with a clear privacy policy and content rules.
- Read the privacy label in the App Store or the Data safety section on Google Play.
- Don’t share identifying details such as your address, workplace, financial info or other people’s private information.
- Use reporting tools when a reply crosses a line.
- Set time limits if you notice the app crowding out sleep, work or people.
- Don’t treat it as medical or mental health advice. AI companions can’t diagnose or treat anything.
- Keep humans in the loop. Friends, family, and professionals when needed.
How Xin approaches safety #
Xin handles the privacy risk through design: its AI model runs on your phone, there’s no account, and your chats and memories stay in a local database on the device. The exceptions, listed in its privacy policy, are ads on the free tier (through Google AdMob), anonymous purchase receipts, and any excerpt you choose to report.
Xin is for adults 18+ only. It asks your age at setup and blocks anyone under 18, every character is an adult, and companions you create must be 18 or older. You can report any reply or companion from inside the app. Its characters are designed to stay in character, which makes for good storytelling but also means Xin, like other companion apps, isn’t a place to seek help in a crisis.
If you’re struggling #
If you’re thinking about suicide or self-harm, or you’re in emotional distress, you can call or text 988 to reach the 988 Suicide & Crisis Lifeline in the US, or chat at 988lifeline.org. It’s free, confidential and available 24/7. In an emergency, call 911. Outside the US, contact your local emergency number or crisis line.
Frequently asked questions #
Are AI companion apps safe for teenagers? #
Most AI companion apps are designed for and rated for adults, and regulators are investigating their effects on minors. Teens are better served by general assistants with teen protections, used with parental guidance. Parents should rely on device-level controls rather than in-app age checks.
Can AI companions be addictive? #
They can encourage heavy use, and research has linked heavier voluntary use with more loneliness and emotional dependence. Setting time limits and watching how you feel after using the app helps.
Do AI companion apps sell your data? #
Practices vary. Many cloud apps share device identifiers and usage data with advertisers and analytics providers, which their store privacy labels disclose. Few openly “sell” chat content, but your chats are usually stored on their servers. On-device apps keep chat content on your phone.
Is it safe to talk to an AI companion about personal problems? #
Talking things through can feel helpful, but an AI can give poor advice and isn’t a substitute for a professional. Avoid sharing identifying details in cloud apps. For anything serious, talk to a doctor or counselor, and for a crisis, call or text 988 in the US.