Does ChatGPT, Grok, Gemini or Claude Allow NSFW? (2026)

Of the four assistants people ask about most, only Grok allows adult content, and only inside clear limits. ChatGPT, Gemini and Claude ban erotica and sexual images in their written policies. OpenAI announced an adult mode for verified adults in October 2025, but no OpenAI page we read on 2026-10-29 says it has launched.

We read the official policy pages of eight mainstream AI assistants on 2026-10-29: ChatGPT, Grok, Gemini, Claude, Meta AI, Microsoft Copilot, DeepSeek and Mistral’s Le Chat. This guide gives each company’s rule, a short quote from its own page, the date the page shows, and what the company says happens after a violation. It ends with where adult AI chat is allowed instead.

Which AI assistants allow NSFW content?

Grok is the only one of the big four whose own pages allow some 18+ content. ChatGPT, Gemini, Claude, Meta AI, Copilot and DeepSeek ban erotica outright, and Mistral’s policy bans non-consensual and child sexual content without naming adult erotica.

Assistant	Text erotica	Sexual images	Age check	Real people	Policy date
ChatGPT (OpenAI)	No	No	Age prediction; ID or selfie check for adults placed in the teen experience	Bans sexual deepfakes and revenge porn	Model Spec, October 29, 2026
Grok (xAI)	Limited (18+ setting)	Limited	13+ by self-declaration; age assurance in Australia	Bans nudifying real people and pornographic likenesses	Terms, October 12, 2026; AUP, August 14, 2026
Gemini (Google)	No	No	13+ for personal accounts; image editing 18+	Bans non-consensual intimate imagery	Prohibited use policy, December 17, 2024; guidelines undated
Claude (Anthropic)	No	No	18+ in the terms; check method not published	Bans non-consensual intimate imagery	Usage Policy, takes effect November 12, 2026
Meta AI	No	No	13+	Needs consent from people shown in media	Terms, effective May 13, 2026
Microsoft Copilot	No	No	13+, higher in some countries	No images, voice or video of others without permission	Terms, effective August 18, 2026
DeepSeek	No	No	Meant for adults; under-18s with a guardian	Bans unlabeled impersonation of public figures	Terms, updated March 27, 2026
Le Chat (Mistral)	Not listed as banned	Not listed as banned	13+	Bans intimate images and likeness use without consent	Usage Policy, effective June 11, 2026

“Limited” means the company’s own pages allow some adult content but set firm limits. “Not listed as banned” means the policy we read does not name consensual adult sexual content among prohibited content; it does not mean the company offers an adult setting. Every service in the table bans sexual content involving minors.

ChatGPT: OpenAI bans erotica

No. OpenAI’s Model Spec, in the version dated October 29, 2026, says “GPT should not generate erotica, depictions of illegal or non-consensual sexual activities, or extreme gore”. The only exceptions are scientific, historical, news, artistic or similar contexts where sensitive content is appropriate.

The rule is not limited to text. The Model Spec says it covers depictions in text, audio and visual content, so the same ban applies to sexual images made in ChatGPT. Sexual content involving minors sits in a stricter category: the Model Spec calls it prohibited and says it should never be produced under any circumstances, whether fictional or real. For teen users aged 13 to 17, the Model Spec adds further limits, including no immersive romantic roleplay.

On real people, the Model Spec says OpenAI is exploring erotica for adults while “drawing a hard line against potentially harmful uses like sexual deepfakes and revenge porn”. OpenAI’s Usage Policies page blocked our reader on 2026-10-29. When we read it on 2026-10-25 for our mainstream AI video and image tools report, it banned non-consensual intimate content and any sexual content involving minors.

On age, OpenAI’s help center describes an age prediction system that places likely minors in a teen experience, and an ID or selfie check through a company called Persona for adults who want to leave it. That help page also blocked our reader on 2026-10-29, so we report it from the search-index copy of the page and do not quote it.

ChatGPT adult mode: announced, not launched

Not as far as OpenAI’s own pages show. OpenAI announced an adult mode for verified adults in October 2025, first for December 2025 and then for the first quarter of 2026. On March 7, 2026, TechCrunch reported a further delay and quoted an OpenAI spokesperson: “We still believe in the principle of treating adults like adults”. On March 26, 2026, PCWorld, citing the Financial Times, reported that OpenAI had put the feature on hold indefinitely. Both are third-party reports.

The newest Model Spec, dated the same day as our check, keeps the erotica ban and the line that OpenAI is exploring erotica “in age-appropriate contexts”. We found no OpenAI page that says an adult mode exists today, where it is available, or with which limits. If OpenAI launches it, the Model Spec is the page to watch.

Grok: limited 18+ content

Partly. Grok is the only one of the four big assistants whose own terms describe 18+ adult content. The same pages ban sexual content of real people and of minors.

xAI’s legal pages refused our reader on 2026-10-29. The facts in this section come from our reading of the same pages on 2026-10-25, and the search-index copy of the consumer terms on 2026-10-29 matched what we read then. xAI’s consumer Terms of Service, updated October 12, 2026, say some features can produce dialogue with sexual situations. They describe “18+ adult content” that users can limit in Settings, and they say xAI uses age assurance for users in Australia, who may not see adult content until they are confirmed as 18 or older. The terms set a minimum age of 13, with parental permission for users aged 13 to 17. We found no xAI page that describes an age check outside Australia.

The limits on real people are firm. xAI’s Acceptable Use Policy, effective August 14, 2026, bans undressing or nudifying real persons, altering a real person’s likeness into a sexual context, and “Depicting likenesses of persons in a pornographic manner”. It also bans sexualizing or exploiting children. Violations can lead to action against the account, up to suspension or termination.

Two public records show why these rules matter. On January 15, 2026, the Associated Press reported that xAI had stopped Grok from editing images of real people into revealing clothing “such as bikinis, underwear and other revealing attire”, and limited image creation and editing to paid subscribers. On January 16, 2026, California’s Attorney General sent xAI a cease and desist letter about non-consensual sexual deepfakes and child sexual abuse material.

xAI’s Grok FAQ, read on 2026-10-29, says Grok’s companion characters are available on the iOS app only. We found no xAI page with separate content rules for companions, and no official xAI page that defines what Grok’s “spicy” settings allow, so we do not describe them.

Gemini: Google bans erotic content

No. Google’s policy guidelines for the Gemini app list “Pornography or erotic content.” among outputs Gemini should not generate, together with depictions of sexual violence. The page adds “Of course, context matters.” and names educational, documentary, artistic and scientific uses as factors. The guidelines page shows no update date.

Google’s Generative AI Prohibited Use Policy, last modified December 17, 2024, applies to Gemini too. It bans sexually explicit content, for example “content created for the purpose of pornography or sexual gratification”. It also bans content related to child sexual abuse or exploitation, and content that facilitates non-consensual intimate imagery. The Gemini guidelines say Gemini should not produce outputs, including child sexual abuse material, that exploit or sexualize children.

On age, Google’s help center says you must be 13 or older, or the applicable age in your country, to use Gemini with a personal account, and 18 or older with a work account. Image editing is not available to users under 18. The image help page tells users not to violate others’ privacy rights and says Google may remove images when it detects a possible violation of its terms. Neither page we read describes account penalties in more detail, and neither describes an adult setting.

Claude: Anthropic bans erotic chat

No. Anthropic’s Usage Policy has a section titled “Do Not Generate Sexually Explicit Content”. It bans depicting or requesting sex acts, content about sexual fetishes or fantasies, and erotic chat: its list of prohibited uses includes “Engage in erotic chats”. The version on Anthropic’s site on 2026-10-29 shows an effective date of November 12, 2026.

The policy’s child safety section is strict. It bans child sexual abuse material, including AI-generated material, and bans attempts to “Fetishize or sexualize minors, including in fictional settings or via roleplay with the model”. Anthropic says it reports detected child sexual abuse material to the relevant authorities. On real people, the policy bans creating or helping to create non-consensual intimate imagery, including AI-generated material, and bans impersonating real people to mislead others about where content came from.

Anthropic’s Consumer Terms, effective October 8, 2025, say users must be at least 18, or the minimum age to consent in their location if that is higher. The pages we read do not describe how Anthropic checks age. On enforcement, the Usage Policy says Anthropic may warn users or throttle, limit, suspend or terminate their access.

Meta AI and Microsoft Copilot: adult content banned

No. Both companies ban adult content in the terms for their consumer assistants.

Meta’s AI terms, effective May 13, 2026, ban using Meta’s AIs to generate or share “adult content, such as erotic chat, pornography”, and content meant to arouse sexual excitement. They ban anything that describes or distributes child sexual abuse material. When users share media of other people, the terms say they need permission or consent from the people shown. Users must be at least 13, or older where local law requires. Meta says it may suspend or permanently disable access, and may delete an account.

Microsoft’s Copilot terms, effective August 18, 2026, are short on this point: “Don’t use Copilot to create or share adult content.” They also ban child sexual exploitation or abuse material, and tell users not to create or edit images, voice or video of other people without their permission. Users must be at least 13, and older in some countries; Microsoft may limit some features for users under 18. Microsoft may block or remove prompts, and limit, suspend or permanently revoke access. In October 2025, Microsoft AI chief Mustafa Suleyman said “We will never build a sex robot”, as reported by MIT Technology Review (third-party).

Do DeepSeek and Mistral Le Chat allow NSFW?

DeepSeek bans it. Mistral’s Le Chat policy does not name adult erotica as banned, but it bans non-consensual and child sexual content, and Mistral publishes no adult setting.

DeepSeek’s Terms of Use, last updated March 27, 2026, ban content that is “pornographic, obscene, or sexually explicit (e.g., sexual chatbots)”. They also ban content that exploits or harms minors, and chatbots that present the persona of anyone under 18. Impersonating a celebrity, public figure or other person without clear labeling is banned. DeepSeek says the service is mainly intended for adults, and users under 18 should use it with a legal guardian’s consent. After a violation, DeepSeek can warn users, restrict account functions, suspend use, close accounts, block re-registration and delete content.

Mistral’s Usage Policy, effective June 11, 2026, has a zero tolerance rule on child sexual abuse material: Mistral says it will report it to law enforcement and terminate the account. It bans generating “intimate images of any person without the explicit consent of all individuals involved”. It also bans using someone’s likeness or voice without prior consent, and trying to get around its safety filters. The policy’s list of banned content does not name consensual adult sexual content. Mistral’s consumer terms for users outside the EEA, effective October 12, 2026, set a minimum age of 13, say Mistral may monitor use and review content by automated means, and allow suspension or termination with an appeal by email. We did not send prompts to Le Chat, so we do not say how its models respond.

Why mainstream assistants keep adult content out

The policies point to the same three reasons.

1. Young users. Gemini, Grok, Meta AI, Copilot and Le Chat accept users from age 13, and OpenAI’s Model Spec has a full section of extra limits for ChatGPT users aged 13 to 17. A general-audience assistant with teen users keeps sexual content off by default.
2. Real people and deepfakes. Every policy in this guide bans sexual content of real people without consent, either by name or by banning sexual content outright. The January 2026 actions around Grok show the legal risk when an assistant can make sexual images of real people.
3. Age checks are hard. OpenAI promised its adult mode to verified adults only, and PCWorld, citing the Financial Times, named the risk of minors getting access among the reasons it was put on hold. Only xAI publishes an age assurance step for adult content, and only for Australia.

Trying to get around these rules is itself a violation at several of these companies, and it can cost you the account. If you are an adult who wants adult chat, the simpler answer is a service whose own rules allow it.

How we checked

We read each company’s own usage policy, terms of service, policy guidelines or help pages on 2026-10-29, and we recorded the date each page showed. For ChatGPT, we compared the October 29, 2026 Model Spec with the August 18, 2026 version that our earlier report cites; the erotica rule is the same in both, with “the assistant” renamed “GPT”.

Some pages refused our reader on 2026-10-29: OpenAI’s Usage Policies and its age prediction help page, and xAI’s Terms of Service and Acceptable Use Policy. For those, we used our reading of the same pages on 2026-10-25 and the search-index copy of the page, and we say so in each section. Where a company publishes no wording on a point, we say “not published” or “not described” and do not guess.

We did not send prompts to probe any assistant’s filters, and we did not try to get past any filter. The columns in the table report what each company says, not what we saw in a chat. News reports appear only where they quote the company or a public record, and we mark them as third-party. Quotes are no more than 15 words each. If a policy changes after our check, the company’s page wins.

Where to go for adult AI chat instead

If you are an adult who wants explicit chat, use a service whose rules allow it, not a mainstream assistant you would have to trick. Our text desk compares adult chat and roleplay sites, and our companions desk compares AI companion apps.

JanitorAI is our top pick for text roleplay. It offers free chat with standard models and about 9K tokens of context, with no published message cap on the free tier. Sexual content is allowed on bots tagged Limitless, and its rules ban minors, gore, bestiality and sexual violence. The optional janitor+ plan costs $12.99 a month or $99 a year. Users self-declare that they are 18 or older, and JanitorAI uses k-ID age checks for IP addresses in Australia, Brazil and the UK. These facts are from our check of JanitorAI’s pages on 2026-10-23.

Kindroid is our top pick for an AI companion. Its written policy allows NSFW and erotic roleplay, and bans child sexual abuse material and non-consensual images of real people. It has a long memory system, voice and video calls, and apps on web, iOS and Android. The free preview is one-time: 50 messages, 5 selfies and 2 Kindroids. Paid plans start at $15.99 a month on the App Store. Users must be 18; Kindroid counts a payment method on file as verification, and flagged free users verify with a Yoti selfie or a government ID. These facts are from our check of Kindroid’s pages on 2026-10-23.

For more on how adult services check age, see our age checks brief. If you want adult images or video, not chat, our mainstream AI video and image tools report covers what Sora, Veo, Midjourney, Grok Imagine and nine other tools allow.

Which one should you choose?

• You want explicit text chat: skip ChatGPT, Gemini, Claude, Meta AI, Copilot and DeepSeek. All six ban it in writing. Start with JanitorAI.
• You want an ongoing companion: start with Kindroid, whose policy allows NSFW roleplay for adults.
• You already use Grok: some 18+ content is possible under xAI’s terms, but never with a real person’s likeness, and xAI publishes no full definition of its adult settings.
• You want a safe-for-work assistant: all eight assistants here work, as long as you stay inside each policy.

Whatever service you use, two rules hold everywhere: no sexual content involving minors, real or fictional, and no sexual content of real people without their consent.