Short answer
Candy AI is an adult-policy hosted companion, not a no-rules model. It sells NSFW chat and 18+ images. Its own Blocked Content Policy still forbids minors, CSAM, non-consensual intimate imagery, and deepfakes of real people, and it can review flagged accounts. Deep Mode is not a current documented setting in Candy help. If you want that hosted adult product, start with Candy AI and read the live policy before you pay.
Privacy and retention are a different job: Is Candy AI safe?. Category hub: uncensored chatbots.
Affiliate disclosure: Dating Auditors may earn a commission if you join Candy AI through links on this page. The adult-policy limits below come from Candy’s public legal pages and help center and apply regardless of that relationship.
Evidence used
Checked 31 August 2026. These are operator statements, not a lab of refusal rates, a filter-pass percentage, or a free-vs-paid model bake-off. We did not re-run chats.
- Candy Blocked Content Policy (revised 30 July 2026) — prohibited: illegal content; hate/discrimination; violence and self-harm; CSAM and content resembling minors; privacy/IP infringement; impersonation and celebrity content; non-consensual intimate imagery and deepfakes of actual persons. The page says users cannot upload or share content. Moderation uses “proprietary LLM” controls; flagged content and other account content may be manually reviewed. Conversations are described as “generally confidential,” which is not a no-moderation promise.
- Candy Terms of Service (revised 17 August 2026) — free users may generally send a maximum of 5 messages; further use needs a subscription. Points to the Blocked Content Policy. Zero-tolerance CSAM language and NCMEC reporting.
- Candy help: What can I do as a paid user? — unlimited chatting; 100 tokens/month for image generation, video, custom AIs, and more.
- Candy help: What is Premium Plus? — named benefits include unlimited text, Create My AI, Generate 18+ images, live audio, memory slots (20). A later fetch of this article was Cloudflare-walled; the benefit list matches the public snippet retrieved the same day.
- Candy help: How do I change chat mode or personality settings? — Classic Mode and Roleplay Mode for Create My AI characters only. Switching to Roleplay Mode is permanent for that character. The article does not call either mode uncensored.
- Candy marketing: uncensored AI chatbot and NSFW AI chat — sell “no filters,” “talk about anything,” and (on the uncensored page) E2EE plus “data is not stored long-term.” Those marketing lines conflict with the legal pages already reviewed on the safety page.
- A site search of
candy.aiandeverai.zendesk.comon 31 August 2026 did not return a current help article or product setting named Deep Mode. Older Dating Auditors copy used that brand; treat it as stale unless Candy republishes it.
What “uncensored” means here — and what it does not
| Claim you will see | What public Candy documents support | Label |
|---|---|---|
| Adult / NSFW chat is the product | Marketing pages sell uncensored and NSFW chat. Premium Plus help names Generate 18+ images | Operator-stated |
| No rules, no refusals, anything goes | Blocked Content Policy lists hard bans (minors, CSAM, NCI/deepfakes of real people, hate, illegal). Moderation can review flagged chats | Marketing vs policy — follow the policy |
| Deep Mode is the adult switch | Not found as a current help setting. Documented CMAI toggle is Classic vs Roleplay, and Roleplay is not described as an uncensored key | Unknown / stale brand |
| Free is a filtered model; paid is open | Terms: free is generally 5 messages, then pay. Paid help: unlimited chat + tokens. No sourced “safety-wrapped free stack” | Message cap is operator-stated; filter-by-tier is unknown |
| Roleplay Mode = uncensored | Help: CMAI-only, switch is permanent. No adult-policy claim on that page | Do not treat as the old Deep Mode |
Do not jailbreak Character.AI or Replika to “unlock” Candy. Candy is a different host with its own policy. Mainstream-filter status: Replika and Character.AI.
Free vs paid (what is documented)
The previous version of this page treated free-tier refusals as a separate “safer stack” and Premium Deep Mode as the real product. We do not have operator text for that routing story.
What we do have:
- Free — generally five messages, then a paywall (Terms).
- Paid — unlimited chatting; 100 tokens/month for extras (images, video, custom AIs, voice). Chat itself is not token-metered once you are paid, per the extra-tokens help article.
- Premium Plus — the help article that names Generate 18+ images as a benefit. Whether every lower paid plan can generate 18+ stills is unknown here; confirm the live subscription screen.
Whether a given explicit prompt is refused, blurred, or billed is unknown without an account test. This page is not that test.
Primary pick: Candy AI
Choose Candy if you want a hosted adult girlfriend chat (no user-pasted OpenRouter key), you accept cloud processing, and you can live inside the Blocked Content Policy — fictional adult RP, not real-person likeness, not minors.
Do not choose Candy if you need:
- a documented zero-moderation or zero-log guarantee;
- deepfakes or NCI of real people;
- “Deep Mode” as a labeled in-product switch (it is not in current help);
- a verified filter-pass score from Dating Auditors.
If the job is a large community character catalog on a phone, CrushOn is the catalog path (official website). If the job is voice or in-chat media, that is Muah with the limits on the multimodal privacy page.
CTA: Open Candy AI — read Blocked Content Policy in-account
Before you renew: confirm the plan that actually lists 18+ images if stills are the reason you are paying, and do not treat marketing “uncensored” as stronger than the legal page. Deletion and logs: account deletion.
Candy AI — adult product with a written Blocked Content Policy, not Deep Mode