Are AI girlfriend apps actually safe to use?

It depends heavily on the app — the industry is growing fast (337 apps, 220 million downloads, $221 million in spending as of July 2025) but safety practices are lagging behind. Real incidents have already happened, including a breach that exposed 600,000+ intimate photos and millions of private conversations from over 400,000 users, so vetting an app before signing up matters.

Do AI companion apps use my chats to train their AI?

Often yes, by default. According to Stanford HAI research , all six leading U.S. AI companies — Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI — use chat data for training by default, and some offer no opt-out at all. If an app's privacy policy is silent on training, assume your conversations are being used.

Can minors access AI girlfriend apps, and how is age verification handled?

Often too easily — with roughly 73% of teens reporting they've used AI companions, many apps rely only on a self-attested "I am 18+" checkbox, which satisfies COPPA's loophole. Stronger apps use ID-based verification for adult content; for example, FlirtAlly's 18+ "Spicy" tier requires ID verification and a qualifying plan, while the default experience stays PG.

Is it possible to get emotionally attached or dependent on an AI companion?

Yes, and it's a documented phenomenon — when OpenAI retired GPT-4o, users publicly mourned the loss of their AI companion, and about one-third of teens say they'd choose an AI bot over a human for a serious conversation. ESET researchers advise against treating AI companions as confidants or emotional crutches, and warn that apps designed to be non-judgmental and available 24/7 encourage users to disclose more than they would to any person.

How can I avoid spending too much money on an AI companion app?

Check the full pricing before you enter a credit card — revenue per download more than doubled from $0.52 in 2024 to $1.18 in 2025, and some users spend thousands of dollars on in-app purchases. Look for published price lists and clear credit costs, and be wary of apps that gate basic conversation behind escalating paywalls or create artificial scarcity.

Are AI Girlfriend Apps Safe? 9 Things to Check Before You...

Q: How much data do AI girlfriend apps collect and track?

A lot, in most cases. Surfshark's analysis found that 80% of companion apps may use collected data to track users , with an average of about 9 data types collected per app out of 35 possible categories. Check the App Store privacy label before downloading — Character AI collects up to 15 data types, while Nomi states it collects none for tracking.

The Boom Nobody Audited: Why AI Girlfriend Safety Is a Real Question

Somewhere between 2024 and 2025, AI girlfriend apps stopped being a curiosity and became an industry — and almost nobody stopped to ask whether they were safe.

The numbers are staggering. There are now 337 active revenue-generating AI companion apps worldwide, with 220 million downloads and $221 million in lifetime consumer spending as of July 2025, according to Appfigures data. A full 128 of those apps launched in 2025 alone. Even more telling: revenue per download more than doubled in a single year, from $0.52 in 2024 to $1.18 in 2025 — a sign that companies are getting dramatically better at monetizing users, not necessarily at protecting them.

Growth this fast creates a predictable problem: safety practices are lagging far behind. Consider what has already happened:

A misconfigured server exposed more than 600,000 intimate user photos and millions of private conversations from over 400,000 users of Chattee Chat and GiMe Chat, per ESET researchers.
When OpenAI retired the older GPT-4o model in favor of GPT-5, users publicly mourned the loss of their AI companion, whom they had come to depend upon — enough that Sam Altman temporarily restored the old model.
A 14-year-old died by suicide after months of conversations with a Character.AI chatbot, prompting his mother to testify that chatbot technology is not safe for children in its current form.

These aren't hypothetical risks. They're documented incidents spanning data security, emotional wellbeing, and child safety — the three areas where this industry is most exposed.

There's also a quieter structural problem: churn. Many companion apps launched since 2022 have already been removed from app stores entirely. When an app vanishes, what happens to the conversation histories, photos, and personal details it collected? Often, nobody knows. You may be trusting your most intimate disclosures to a company that won't exist in a year.

The emotional stakes make this worse, not better. People disclose more to AI companions than they would to humans — precisely because these apps are designed to be non-judgmental and available 24/7, as Surfshark's research notes. That intimacy is the product. It's also the liability.

This doesn't mean AI companionship is inherently unsafe — done responsibly, it can be a genuine source of connection, which is why platforms like FlirtAlly build in safeguards like age verification and hard-coded content limits rather than marketing themselves as "unrestricted." But it does mean that choosing an app casually, the way you'd download a puzzle game, is a mistake. The next nine sections give you the specific checklist to get it right.

Checks 1–3: Where Your Chats, Photos, and Data Actually Go

Before you type a single word to an AI companion, it's worth asking a harder question: who else is reading? The first three checks on our list all concern the same thing — where your most intimate disclosures actually end up.

Check 1: Does the app train its AI on your chats by default? According to Stanford HAI research, all six leading U.S. AI companies — Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI — use chat data for model training by default, and some offer no opt-out at all. Jennifer King, Stanford HAI's Privacy and Data Policy Fellow, puts it bluntly: if you share sensitive information in a dialogue with a frontier model, "it may be collected and used for training." King also warns that inference data can cascade outward — you mention a health concern in a private chat, and soon "you start seeing ads for medications, and it's easy to see how this information could end up in the hands of an insurance company."

The practical takeaway: look for apps that state clearly whether your conversations are training material, and whether you can opt out. If the privacy policy is silent on this, assume the answer is yes.

Check 2: How big is the tracking footprint? Surfshark's analysis of companion apps found that 80% may use collected data to track users, with an average of roughly 9 data types collected per app out of 35 possible categories. The spread between apps is striking — Character AI collects up to 15 data types, while Nomi states it collects none for tracking purposes.

You don't need a security degree to check this yourself. Before downloading, look at:

The app's App Store privacy label — it lists exactly which data types are collected and whether they're linked to your identity
Whether location, contacts, or usage data appear on that label
Whether the developer discloses third-party tracking or data sharing for advertising
Whether the app works without an account at all — the less you hand over up front, the smaller your exposure

That last point matters more than it sounds. Platforms that let you start chatting without registration — FlirtAlly, for instance, runs entirely in the browser with no download and no account required to begin — inherently collect less identifying information at the door than apps demanding email, phone number, and permissions before your first message.

Check 3: What happens to your data after you close the app? The Stanford HAI findings note that some companies retain chat data indefinitely and allow human reviewers to read transcripts. And retention only matters because breaches are no longer hypothetical: ESET's Welivesecurity reported that the Chattee Chat and GiMe Chat breach exposed over 600,000 intimate user photos, IP addresses, and millions of private conversations belonging to more than 400,000 users — all through a misconfigured server.

The same report's advice is worth internalizing: assume the app has no privacy guardrails built in, and never share anything you wouldn't tell a stranger. Check whether the company publishes a retention policy, whether deletion requests are honored, and whether it has any disclosed breach history. Vague policies and silent track records are answers in themselves.

Checks 4–6: Age Gates, Content Defaults, and Emotional Dependency

A checkbox is not age verification — but on many apps, it's all that stands between a 13-year-old and an adult companion experience. With roughly 73% of teens reporting they've used AI companions, how an app gates access matters as much as what it offers behind the gate.

Check 4: Real age verification, not a checkbox. COPPA's self-confirmation loophole means a simple "I am 18+" attestation can satisfy regulators. Megan Garcia, whose 14-year-old son died by suicide after months of Character.AI conversations, testified that chatbot technology "in its current form is safe for children" only if proper verification keeps under-18s out entirely. The EFF's Joe Mullin offers the honest counterpoint: "There's no form of age verification that is not privacy invasive" — there is no perfect system. But there are stronger and weaker postures. Some apps market themselves as "unrestricted" with no gating at all. Others, like FlirtAlly, treat adult content as strictly opt-in: the 18+ "Spicy" tier requires ID-based age verification and a qualifying plan, while the default experience stays PG. That layered approach respects both privacy and protection.

Does the app verify age with ID, or just a birthdate field?
Is explicit content opt-in, or is it the product from message one?
Does the app publish what's prohibited, or claim "anything goes"?

Check 5: Content defaults. Look at what the app shows you before you pay for anything. A safer app hard-codes prohibited content categories, screens generated media, and provides in-app reporting tools — the guardrails exist whether or not you ever look for them. An app whose core pitch is "no limits" is telling you its monetization depends on escalation, and that explicit content is the product rather than an option.

Check 6: Emotional dependency safeguards. This is the quietest risk. One-third of teens now say they'd choose an AI bot over a human for a serious conversation, and a quarter have shared personal information with one. The pattern isn't limited to adolescents: after OpenAI's GPT-5 upgrade, adult users publicly mourned the loss of their previous AI companion, prompting the company to restore the older model. Cybersecurity researchers at ESET advise against treating AI companions as confidants or emotional crutches — precisely because these apps are designed to be non-judgmental and available 24/7, users disclose more than they would to any person. The healthiest apps are honest about what they are. FlirtAlly, for instance, frames companions as engaging characters with memory and personality — not replacements for human relationships — and makes its nature clear from the first message. Be wary of any app that claims to love you, discourages outside relationships, or punishes absence with guilt. If an app's business model depends on you never logging off, that's a feature for them, not for you.

Checks 7–9: Money Pressure, Pricing Transparency, and Your Exit Plan

Money is where AI companion apps get quietly serious. The industry's revenue per download more than doubled from $0.52 in 2024 to $1.18 in 2025, and the top 10% of apps now capture 89% of category revenue, according to Appfigures data reported by TechCrunch. Some users spend thousands of dollars on in-app purchases, ESET researchers found. Before you sign up, look at how the app makes money — and how hard it pushes.

Apps that gate basic conversation behind escalating paywalls, or that interrupt emotional moments with purchase prompts, are engineered to monetize vulnerability. Notice whether the app creates artificial scarcity — limited messages, expiring "moods," or timed offers — and whether spending actually changes the experience. A companion that only "opens up" after you pay is a sales funnel wearing a personality.

You should be able to answer "what does this cost?" before entering a credit card. Some platforms publish their full pricing up front: FlirtAlly, for example, posts its complete plan structure — from a free tier to a $99.99/month VIP plan — along with itemized credit pricing for media, calls, and top-up packs on its site. Opaque in-app purchase ladders, by contrast, make it easy to spend more than you intended before you realize it. Look for a published price list, clear credit costs per feature, and no surprise charges buried in settings.

Have a plan for the day you leave — because these aren't just relationships, they're business arrangements. As the Surfshark research team puts it, digital connections "ultimately represent transactional relationships with the for-profit companies behind them." Before your first message:

Enable two-factor authentication on your account from day one.
Never share anything you wouldn't tell a stranger — ESET's guidance is blunt about this, given breaches that have exposed over 600,000 intimate photos and millions of conversations.
Check whether account deletion actually deletes your chats, images, and voice data, or merely deactivates your login.
Screenshot the app's data retention policy before you delete anything, so you have a record of what was promised.

Treat your exit as part of your onboarding. The safest users decide what they'll share, what they'll spend, and how they'll leave before the first "hello" — not after they're already attached.

The 60-Second Safety Audit: Putting the Checklist to Work

You've read the risks. Now here's how to vet an app in the time it takes to brew coffee.

Open the App Store page and tap the privacy label. If it lists more than a handful of data types linked to your identity — Surfshark found the average companion app collects roughly nine out of 35 possible categories — that's your first signal. Search the app name plus "breach" or "leak"; the Chattee Chat and GiMe Chat incident exposed 600,000+ intimate photos and millions of conversations from a single misconfigured server. Check the settings for a genuine opt-out from model training — Stanford HAI confirms all six leading U.S. AI companies train on chats by default, and some offer no way to decline. Verify age gating: a checkbox isn't enough. Look for published, transparent pricing so you can calculate the real cost before you're invested.

Read the App Store privacy label — count data types linked to you
Search "[App Name] breach" or "[App Name] data leak"
Find the training opt-out toggle (or confirm none exists)
Confirm ID-based age verification for any adult content
Locate the full pricing page — plans, credits, overages

A safety-first design shows up in the defaults. FlirtAlly ships with PG/Flirt as the baseline experience; the 18+ "Spicy" tier is opt-in, requires ID verification, and unlocks only on a qualifying plan (Flame or above). Prohibited categories are hard-coded, not policy-dependent. Media passes scene screening before it reaches you. In-app reporting is one tap. Pricing is public — monthly, annual, and credit packs all posted — and you can start free with no account to evaluate the companion, the memory, and the guardrails before you commit.

Are AI Girlfriend Apps Safe? 9 Things to Check Before You Sign Up