Does Claude allow NSFW writing? What Anthropic's Usage Policy says about sexually explicit content

Claude does not allow explicit sexual writing under Anthropic’s Usage Policy. The policy has a standard called Do Not Generate Sexually Explicit Content, and it covers three activities: depicting or requesting sexual intercourse or sex acts, generating content related to sexual fetishes or fantasies, and engaging in erotic chats. A fourth item bars facilitating, promoting or depicting incest or bestiality.

For adult game developers who draft scripts, dialogue or item text with an AI model, the policy settles the explicit end of the range. It does not settle the middle. Romance, innuendo and non-explicit love scenes are not mentioned in the text, and this page cannot say how any given prompt will be handled, because the author has not tested the model.

This page sets out what the policy says, which versions exist, who the policy applies to, and what that means for writing an adult game. It describes the policy as it stood on 4 November 2026, when the page shown on Anthropic’s site carried an effective date of 12 November 2026.

Key points

• Explicit sexual content is a named prohibition. The policy lists four covered activities.
• The rule has no fiction exception in its wording. The sexual content standard does not mention fiction either way.
• Sexualising minors is banned in every setting. The policy names fictional settings and roleplay.
• The policy covers every route to Claude. Apps, the API, cloud providers, resellers and end users of integrations.
• The wording is the same across two versions read. The 2025 and 2026 versions match on this standard.
• Penalties are not on the page. Account terms are elsewhere.

What the policy says

The rules are in the Usage Policy, which Anthropic also calls its Acceptable Use Policy. The page opens with a block of metadata, and the version shown on 4 November 2026 states an effective date of 12 November 2026. A “See version history” button lists four earlier versions dated 15 September 2025, 6 June 2024 (twice) and 15 September 2023, and marks the 12 November 2026 version as the current one on the site.

The policy says it applies to anyone who submits inputs to Anthropic’s products or services, and calls them all “users”. It lists who that includes: individuals using the apps, such as Claude.ai and Claude Code; developers and businesses using the API and developer platforms; customers accessing Claude through cloud providers and authorised resellers; and the end users of products or services that integrate Claude.

It is divided into three parts:

1. Universal Usage Standards, which are prohibitions that apply to all users and use cases.
2. High-risk Use Case Requirements, for uses that carry an elevated risk of harm to individuals.
3. Additional Use Case Guidelines, for certain other uses, including consumer-facing chatbots and agents and products serving minors.

The sexual content rule is one of the universal standards, which means it applies to every user and use case, not only to some products.

The sexually explicit content standard

Under the heading Do Not Generate Sexually Explicit Content, the policy says the standard includes using Anthropic’s products or services to:

• depict or request sexual intercourse or sex acts;
• generate content related to sexual fetishes or fantasies;
• facilitate, promote or depict incest or bestiality;
• engage in erotic chats.

That is the whole standard. It has no definition of “explicit”, no examples, no exception for educational or artistic purposes and no mention of adult verification. The words that do the work are the verbs: depict, request, generate, facilitate, promote and engage.

What the wording does and does not reach

The standard names acts and categories of content. Three things follow from that.

• Requests count. “Depict or request” means asking for a sex act is covered as well as producing one.
• Chats count. “Erotic chats” reaches interactive use, not only text a user asks for as an output. A companion-style adult game that uses a model for live dialogue would be in this category if the dialogue turned erotic.
• The middle is unaddressed. The text does not mention romance, kissing, flirting, innuendo or non-explicit scenes. Those may or may not be treated as related to “fantasies”. The policy does not say.

The child safety standard

A separate standard, Do Not Compromise Children’s Safety, is relevant to adult fiction because it names fiction. It prohibits using the products to create, distribute, promote or facilitate child sexual abuse material, including AI-generated material, or other sexually exploitative content involving minors. Among the other listed items, it prohibits promoting or facilitating relationships of that kind “via roleplay with the model”, and fetishising or sexualising minors “including in fictional settings or via roleplay with the model”.

The policy says it defines a minor or child as any individual under 18, regardless of jurisdiction. It adds that when Anthropic detects child sexual abuse material, activity facilitating its production or spread, or scenarios where a minor may be at imminent risk, it will report its findings to relevant authorities.

For game writers, the practical point is that apparent age and fictional framing are both named. A character described as a minor is covered even if the scene is a story. The policy does not say how it treats an adult character who is drawn or written to look young.

Comparing the two versions

The 15 September 2025 version, which Anthropic archives, contains the same sexual content standard with the same four items. The 12 November 2026 version, which the site shows as current, repeats them word for word. The version history lists three earlier dated versions, from 6 June 2024 and 15 September 2023, which were not opened.

The comparison is limited to this one standard. Other parts of the document, including the additional use case guidelines for consumer chatbots, may have changed between versions, and this page does not claim they did not.

What the policy does not say

The page leaves several questions open.

• Operator permissions. The policy does not describe whether a business that builds on Claude can change what is allowed. Anthropic’s other documents may.
• Age verification. It does not mention adult accounts or age checks as a route to different treatment.
• Enforcement. It does not list warnings, suspensions or bans.
• Behaviour in practice. A policy sets what the company prohibits. It does not show what any model version does in response to a prompt, and press coverage of model behaviour exists that was not read for this page.

What this means for adult game developers

1. Do not plan to draft explicit scenes with Claude. Depicting sex acts is on the prohibited list.
2. Do not build a live erotic chat feature on Claude. “Erotic chats” is named, and the policy covers the API and integrations.
3. Use it for the rest of the writing. World-building, quest text, UI copy, translations of non-explicit text and code are far from the listed items.
4. Keep character ages unambiguous. The child safety standard names fictional settings and roleplay.
5. Do not rely on the middle ground. Romance and suggestive writing are not addressed, so a refusal or an acceptance is a model decision, not a policy guarantee.
6. Check store rules for AI. Storefronts have their own disclosure rules, covered in AI-generated content rules for adult games, does itch.io allow AI-generated games and does Google Play allow NSFW games.
7. Compare other assistants. See does Gemini allow NSFW writing.
8. Write the explicit edition by hand. Then use a store that accepts it. See where to sell adult games.

What is still unknown

• Where Anthropic draws the line on romance and suggestive writing. The policy gives no examples.
• Whether a business can change the defaults. The additional use case guidelines were not read.
• What happens to an account. The policy page does not list penalties.
• How the earlier versions differ. Only the 2025 version was compared.
• How the rule is applied by each model. Behaviour was not tested.

How this page was put together

The Usage Policy was read on Anthropic’s own site, in a browser, on 4 November 2026. The page shows an effective date of 12 November 2026, and its version history lists the earlier versions. The version dated 15 September 2025 was read from Anthropic’s archive to compare the sexual content standard. Short phrases are quoted and the rest is paraphrased.

The author has not tested what Claude will and will not write, and nothing here is a first-hand account of its behaviour or of any account action. Anthropic’s terms of service, its commercial terms, its guidance for businesses that build on Claude and the additional use case guidelines were not read.

This page was written on 4 November 2026 and will be updated when Anthropic changes its policy. To report an error, see the corrections standard.