Candy AI Videos: What You Get, What They Cost in Tokens
Candy.ai sells three kinds of video, and none of them are free once your monthly tokens run out. Here is how each type works, what the tokens buy,...
Candy AI is a real company with written terms, but safety has two sides: your data and your wallet. Here's what the policies actually say.
For most adults, yes. Candy AI is run by a registered company, it has published terms and a privacy notice, and there is no public record of a data breach as of October 2026. But safe does not mean private in the way you might assume. There is no end-to-end encryption, your chats may be used to train models, and refunds are narrow. The bigger day-to-day risks are money and data, not malware or outright fraud.
If you want the full breakdown of features and pricing, our Candy.ai review covers that. This page is about safety: who runs the site, what happens to your data, what you get billed, and how to spot the scams that use the Candy AI name.
Candy AI is operated by EverAI Limited, a company registered in Malta under number C107181, with an office in Santa Venera. Maltese law governs the contract, and EverAI is the data controller. The privacy notice cites GDPR and CCPA as applicable law.
That is a real corporate footprint. It does not mean a regulator has audited how the company handles data. No such audit is published, so compliance is a company claim. Still, a named company with a registration number and a physical address is a different category from an anonymous site.
If you see a charge from EverAI on your statement, that is the legal billing name, not fraud. More on that below.
| Question | Answer |
|---|---|
| Real company | EverAI Limited, Malta, C107181 |
| Scam | No, but refunds strict |
| End-to-end encrypted | No, transit only |
| Training use | Yes, de-identified, no opt-out |
| Retention | Up to 3 years after last activity |
| Statement name | EverAI |
Traffic to Candy AI is encrypted in transit. The site uses HTTPS over TLS 1.3, according to an October 2026 check. Reviews sometimes claim AES-256 encryption at rest, but the privacy notice only says the company uses appropriate security measures. There is no end-to-end encryption. Chats and images are stored server-side, which means the company can access them.
The notice says prompts and outputs may be de-identified and used to train AI models and improve moderation. There is no described opt-out from that training use. Flagged messages may also get manual review by authorized partners. Outside LLM providers, hosting companies and moderation vendors may receive message content.
The company says it will not sell, trade or transfer personal data without consent. That is a meaningful promise, but it sits alongside a broad license in the terms. Section 4 grants a royalty-free, sub-licensable, worldwide, perpetual license on what you input. The company also says it never claims ownership of your content. The two statements can coexist: you keep ownership, but the platform gets broad rights to use what you put in.
Practical takeaway: treat every chat as something a stranger could read. Use a pseudonym. Keep out real names, addresses, workplace details, and anything you would not want in a support ticket.
Data is kept up to three years after your last activity, possibly one year in some cases. Financial records are kept for 10 years for legal reasons, and log files for 30 days.
Resetting a chat clears one conversation, its history and its memory. That is not the same as deleting your data. Account deletion happens in Settings under Danger Zone. Under GDPR you also have rights to access, rectification, erasure, portability and objection. The legal bases for processing are consent, contract, legal obligation and legitimate interest.
If you want erasure, email [email protected] and keep a copy of the request. Deleting chats does not guarantee immediate full erasure, and payment records stay for legal reasons.
Subscriptions auto-renew until you cancel in settings. After you cancel, premium access continues until the period ends. Each subscription includes only 100 tokens a month. Image and video generation spends tokens beyond the plan price, so a cheap monthly price can hide a token meter.
Refunds are only available within 24 hours and only if you have used 20 tokens or fewer. Card payments only. There are no refunds for user-side technical issues or moderated content. In the EU and UK, the 14-day withdrawal right applies minus usage.
Charges appear as EverAI. An unfamiliar EverAI charge is the most common false alarm. A second charge after subscribing is usually media tokens, not a duplicate subscription.
If you want to understand how tokens work before paying, our guide to Candy AI video costs breaks it down. And if you are comparing subscriptions, Candy AI vs OurDream AI puts the pricing side by side.
Candy AI is 18+ only. The age gate is self-declared, with third-party age estimation by location. The company states zero tolerance for underage content and reports known CSAM to NCMEC. An automated filter can flag messages. Explicit chat and fictional adult images are allowed. Real person likeness is banned.
That filter is not perfect, and it can flag legitimate messages. If your content is moderated, refunds are not available for that reason.
| Looks like scam | Reality |
|---|---|
| Unknown EverAI charge | Legal billing name |
| Charged again after subscribing | Media costs tokens |
| Refund denied | 24-hour, 20-token rule |
| APK or app link | Not official; web only |
| Support asks for gift cards | Real scam |
There is no official app or APK. Candy AI is web only at candy.ai. Any app link or APK is not official and should be treated as unsafe. Official channels are the help center, Discord, X, Instagram, TikTok and YouTube. If someone claiming to be support asks for gift cards, that is a real scam. No legitimate support team does that.
There is also the Trustpilot problem. Candy AI's Trustpilot profile was removed for guideline violations. It was around 3.8 out of 5 before removal. Complaints tend to center on refunds and token costs, not on stolen data.
Secrets AI promises AES-256 encryption at rest and immediate chat deletion when you delete your account. OurDream AI advertises end-to-end encryption, but its policy describes server-side encryption. Neither claim replaces reading the policy yourself, but both are more specific than Candy AI's "appropriate security measures" language. For a deeper comparison, see Secrets AI vs Candy AI.
Candy AI's strength is not privacy. It is image generation and character customization, with images, video and voice. We rate it 9.8 out of 10 overall. If pictures are the point, it is a reasonable pick. If privacy is the point, look elsewhere.
This is an adult product built for companionship and fantasy. That is fine. But it is worth being honest with yourself about how much time and money you want to put in. The token meter makes it easy to spend more than you planned. The memory is limited, so the app may forget things you told it. Neither of those is a safety flaw in the security sense, but they affect how the product feels day to day.
If you are using it to fill a gap rather than for fun, set a budget and a time limit before you subscribe. The auto-renewal will not remind you.
Candy AI is a real, regulated platform, and it is reasonably safe for adults who protect their identity. The weak spots are privacy and money. There is no end-to-end encryption, training use has no opt-out, outside providers see message content, retention runs up to three years, and refunds are strict.
Use a pseudonym, keep sensitive details out, and watch the renewal date. If you want alternatives with different privacy or pricing trade-offs, our Candy AI alternatives guide matches them to the complaint you actually have.
For most adults, yes, provided you use a pseudonym and keep sensitive details out of chats. It is a registered business with written terms, but there is no end-to-end encryption and chats may be used to train models. Treat every message as something a stranger could read.
It is legit. Candy AI is operated by EverAI Limited, registered in Malta under C107181, and it delivers the features it advertises. Most scam claims trace back to token costs and the strict refund policy rather than fraud.
Possibly. Flagged content may get manual review by authorized partners, and de-identified chats may be reviewed for training data. There is no end-to-end encryption, so chats are stored server-side.
Delete your account in Settings under Danger Zone, then email [email protected] to request erasure. Keep a copy of the request. Payment records are kept for legal reasons and will not be deleted immediately.
Realistic AI companions with strong memory, video and voice calls