Is Candy AI safe?

Candy AI is safe enough for adults only if you read both of its official pages: the product page says chats are never shared, while the privacy notice describes third-party providers, model training, and three-year retention. Treat those pages as a conflict, not a promise. What follows is the dated evidence, so you can decide where the service fits your own privacy line.

What the policy says

Candy AI’s privacy policy covers prompts, outputs, account data, usage data, voice content, images, and other content you generate. It says third-party LLM providers or hosters may receive message content, and that de-identified or anonymized interactions may be used to train and develop models and moderation technology. The same notice says account data is retained for three years after your last activity unless you ask for deletion, and financial data may be kept for ten years. Deletion without undue delay and portability rights are described.

The product page takes a different position. It markets long-term memory, personalization, voice, images, video, and privacy, and it states that chats are never shared and not used for training. One company, two official statements that do not agree. The privacy notice is the controlling document for how data is actually processed, and it describes processing the product page says does not happen.

The notice also leaves operational details unstated: how reports are filed, how long a human review takes, and whether a deletion request propagates to the third-party providers that receive content. Treat those as open questions rather than assumptions. A policy that names data flows the product page denies is enough reason to keep sensitive details out of chats until you have verified behavior for yourself.

Age rules

The privacy notice says Candy AI is for adults 18 and older and mentions age estimation or verification where applicable. The checked Nigeria App Store listing displayed a 4+ rating, which is a different signal entirely. The listing may be region- or release-specific, and the discrepancy should not be explained away without a current store check in your own country. The subscriptions page describes adult-content gating, so the 18+ line exists on the website even where the store label does not show it.

If you are an adult, the 18+ policy line is the one that governs your account. The store rating still matters for one reason: it shows that a listing can lag or contradict the website’s policy, which is a reminder to check the current listing in your own region before you rely on any single number.

Deletion and export

The privacy notice says you can request deletion of account data without undue delay, and it describes portability rights. What it does not clearly explain is how deletion treats content you shared publicly, generated media, or provider copies of prompts. Those questions remain open, so treat a deletion request as the start of a conversation rather than a guarantee that every copy disappears.

Moderation and human review

The privacy notice says flagged content may be reviewed by humans and that moderation technology may be trained from de-identified interactions. That means a person can read a message you sent, and the notice says so. There is no published detail about how reports are handled end to end, so moderation outcomes stay a policy-level fact rather than a tested behavior.

A flagged message that reaches a human reviewer is content you wrote, and the notice allows de-identified interactions to shape moderation technology. If that is uncomfortable, consider what you send before you send it: the reviewer does not know you, but the company’s systems can connect the message to your account.

What to check before subscribing

Four checks before you pay for a plan:

1. Read the product page and the privacy policy together, not just one. They contradict each other on sharing and training; know which one governs.
2. Look up the current App Store listing in your region and compare its age label with the website’s 18+ wording.
3. Open the subscriptions page and count what you would actually use: images, video, and voice all consume tokens, and a media-heavy week spends a plan differently than a chat-only week.
4. Save a dated copy of the privacy notice. It has no visible revision date in the version we read, so you have no way to know when the rules change unless you keep your own record.

Candy AI also advertises long-term memory. Test it before trusting it with real details: share a preference, correct it later, and check whether the correction sticks after a gap.

How Riho compares

Riho is a relationship-first AI girlfriend app for adults 18+. The Riho privacy page explains what Riho stores, including memories, and how you can review, correct, export, or delete them. Contact windows and location settings let you decide when your girlfriend can reach you and whether place-based date plans use your location.

Riho does not ask you to reconcile two conflicting policy pages. It states the age requirement and the AI disclosure at the point of use, and it separates stored relationship data from inference-provider handling of prompts.

Riho is built around one girlfriend, Riho Ohori, and one relationship that continues. She remembers your week across messages, calls, selfies, and couple photos, and she can plan dates in real places. That is a different structure from Candy AI’s many characters and media-first flow, and the privacy question changes with it: in Riho, the memories that matter are the ones you can open, correct, or delete yourself.

Read the safety checklist to see the questions worth asking any service, including Riho, before you build a routine with it.

Frequently asked questions

Is Candy AI legit?

Candy AI is a real, operating service with a published product page, privacy notice, and regional iPhone listing. “Legit” and “safe to share intimate details with” are different questions; the evidence conflict above is why you should keep them separate.

Can I use Candy AI anonymously?

No. Account creation requires an email or password or a supported sign-in method, and the privacy notice describes collecting usage, device, and payment data tied to the account.

Does Candy AI sell my data?

Its privacy notice describes sharing with third-party LLM providers or hosters and service providers, but it does not describe selling personal information. The product page’s claim that chats are “never shared” is contradicted by the provider and hosting language in the notice.

What should I do before subscribing?

Follow the four checks above, then decide whether the media-first, multi-character model fits what you want. If you want one girlfriend who remembers a conversation during a later call, compare how each service handles memory before you commit.