Is Candy AI Safe? Privacy, Data & Billing Explained

Candy AI is reasonably safe for adults who treat it as semi-private: you can delete chats and your whole account, retention periods are published, and the card charge shows as "EverAI" rather than "Candy AI". The gaps are that only in-transit encryption is documented, no two-factor authentication is documented, and de-identified chats can be used to train its AI.

There is no public profile, follower list or social feed either, so the most common fear, that someone else will see your activity, is largely handled by the way the product is built.

Safe is not the same as private, though. Everything you send is stored on Candy AI's servers and processed by AI models so characters can reply.

Its privacy notice also says de-identified or anonymized conversations may be used to train its models, with possible human review, and that third-party AI providers it uses may receive the content of your messages.

Most cloud-based AI companions work in a broadly similar way, but it means your chats are not private in the way a note kept offline on your own device would be.

The sensible approach is to treat Candy AI like any other account that holds sensitive information, and avoid typing anything you would never want stored anywhere. The sections below explain what is kept, how billing works, and which settings make the biggest difference.

Candy AI stores your account details, your conversation history, the images and voice messages generated in your chats, the characters you create, and basic technical data about your device and activity.

Most of it is needed for the app to work at all: a companion can't stay in character or remember your history without a record of what you've said.

Its privacy notice publishes retention periods: account data is kept for three years after your last activity (it may be shortened to one year) or until you ask for deletion, financial records for ten years, and logs for 30 days.

Candy AI does not ask for government ID when you sign up. You confirm you are 18 or older (a self-declared age gate; extra checks apply only where local law requires them) and sign up with an email, or with Google, Discord or Twitter (X).

Because you only consume content rather than publish it, your real-world identity stays out of the account unless you add it yourself.

The single most effective privacy habit is simple: keep identifying details out of your conversations entirely. That means your full legal name, home address, workplace and any financial information. The AI doesn't need them, and data that is never stored can never be exposed.

Candy AI conversations are only partly private: data is encrypted in transit (TLS), but Candy AI documents no end-to-end or at-rest encryption for chats, and the service has to read your messages to generate replies.

Its help center says deleted chats are permanently removed, and deleting or resetting a chat also wipes that chat's memory.

It helps to be precise about what that encryption does. Encryption in transit protects your messages from being intercepted between your device and Candy AI's servers.

It does not stop the platform itself from accessing content, which it must do to run its AI models and enforce its rules, and Candy AI does not say how chats are protected once they are stored.

• Protected in transit: connections use TLS 1.2 or newer, according to Candy AI's own FAQ.
• Accessible to the platform and its AI providers: Candy AI can access content to run the service, de-identified chats may be used for training, and third-party AI providers may receive message content.
• Controlled by you: deleting chats and characters, or the whole account, is the most direct way to limit what is stored about you.

A useful rule of thumb: assume anything you type exists on a server until you delete it, and delete whatever you don't want kept. There is no training-specific opt-out; users in the EU and UK can use the general GDPR right to object.

Mostly. The charge shows on your card or bank statement as "EverAI", the company behind Candy AI, so it doesn't say "Candy AI" or hint at adult content. Anyone glancing at your statement sees a company name rather than the service.

Someone who looks up "EverAI" could still connect it to Candy AI, so if that matters to you, use a separate card.

Candy AI's plans, as checked for our review (6 October 2026):

• Tokens: images, voice and video use tokens, and Candy AI doesn't publish a per-item token price. Its terms say remaining tokens expire when your subscription ends.
• Payment methods: credit and debit cards, crypto (including BTC, ETH, USDC and Litecoin), Apple Pay, Google Pay and bank transfer, depending on your country.
• Cancelling: in Settings, choose Unsubscribe and confirm the prompts. Refunds are limited to requests made within 24 hours, with 20 tokens or fewer used.

For extra separation, some users pay with a privacy-focused virtual card or a card they keep only for subscriptions. Prices change, so confirm the current figure at checkout.

The biggest privacy gains on Candy AI come from your own account security: use a dedicated email and a unique password, lock your devices, and log out on shared machines.

Candy AI doesn't document a two-factor authentication option of its own, so if you want a second factor, sign in with a Google account that has two-factor authentication turned on.

1. Sign up with a dedicated email that isn't tied to your real name or your main accounts.
2. Use a strong, unique password. Never reuse one from another site; a password manager makes this effortless.
3. Want a second factor? Use Google sign-in with two-factor authentication enabled on that Google account, since Candy AI documents no 2FA of its own.
4. Lock your phone and computer, and don't stay logged in on shared or family devices.
5. Use a private browser window on any device other people use, and clear your history if local privacy matters to you.
6. Delete old content, such as conversations and characters you no longer use, every so often, or delete the account from Settings when you are done.

None of these steps are exotic. They are the same habits that keep email and banking accounts safe.

Applied here, they cover the realistic threat, which is far more often a partner, roommate or relative finding an open session than a sophisticated hack.

For a broader playbook on keeping adult accounts separate from your identity, see our guide to staying anonymous on adult sites.

Candy AI sits toward the more transparent end of the AI companion category, scoring 7.4 out of 10 for safety in our review, ahead of DreamGF (6.5), Janitor AI (6.0) and Replika (5.5).

It publishes retention periods, offers self-serve chat and account deletion, bans deepfakes and anything involving minors, and has human reviewers decide bans.

It loses points for a self-declared age gate, no documented two-factor authentication or end-to-end encryption, training on de-identified chats, and a January 2025 Bellingcat report that its ads reached the deepfake site MrDeepFakes through its affiliate programme (EverAI said it cut ties with that publisher).

• Moderation is a policy backed by enforcement. Candy AI uses automated moderation plus manual review of flagged content, and bans anything involving minors, non-consent, deepfakes or real people's likenesses. Consistent, enforced rules are a good sign, but no filter is a technical guarantee.
• Bring-your-own-model platforms trade safety differently. Janitor AI is free and text-first, but your chats can pass through a third-party language model or proxy, which changes the privacy picture. Our Janitor AI explainer covers how that works.

No app is universally the safest, only the one whose trade-offs match your priorities. As a rule, a platform that documents its data practices, enforces its rules consistently and gives you real deletion controls deserves more trust than one that stays quiet on all three.

The real risks of using Candy AI are ordinary ones: a data breach, someone opening your session, curiosity about your statement, policy changes and emotional over-reliance. None is unique to Candy AI, and each one can be managed.

Candy AI's own rules also guard against the worst misuse: it is for adults only, and it bans anything involving minors, non-consent, deepfakes or real people's likenesses.

Keep your roleplay inside the same lines, with fictional adult characters and nothing modelled on someone you actually know.

Put simply, the strongest protections are the ones you control: share little, secure the account and clean up now and then. Do that, and the remaining risk is low and comparable to the mainstream subscriptions you already use.