Veilwright Studio
Industry

Does Google Veo allow NSFW video? Safety filters, the sexual content category and the Prohibited Use Policy

Google Veo does not allow sexually explicit video. Google’s Generative AI Prohibited Use Policy bars sexually explicit content, giving content created for pornography or sexual gratification as the example, and the Google Cloud documentation for Veo says prompts are checked against safety filters that include a sexual category defined as “sexual or suggestive content”. The category is wider than the policy’s example.

For adult game developers, AI video is a tool for trailers, animated backgrounds and cut-scenes. The question is whether it can be used for anything near the line. The documents say that the filters cover suggestive material as well as explicit, that circumvention is a violation in its own right, and that artistic context is an exception the policy leaves to Google.

This page sets out what the Veo responsible AI page, the Prohibited Use Policy and the Gemini app guidelines say, and what that means for a game project. It describes the pages as they stood on 7 November 2026. The Google Cloud page was machine-translated into Portuguese when read, and the Prohibited Use Policy and Gemini pages were read in Portuguese on 31 October 2026, so the wording below is translated.

Key points

  • Sexually explicit content is prohibited. The Prohibited Use Policy gives pornography and sexual gratification as the example.
  • Veo’s filter category is broader. It covers sexual or suggestive content.
  • Inputs are screened. Text, image and video prompts are evaluated against safety filters.
  • Outputs can be blocked. Fewer videos than requested means some were blocked.
  • Children and celebrities have their own categories. Both are restricted unless a setting or allowlist applies.
  • Evading the filters is prohibited. So is manipulating the model.

What the Veo documentation says

The Google Cloud page Responsible AI for Veo video generation shows a last updated date of 7 November 2026. It describes Veo as designed with Google’s AI Principles and says developers need to understand how to test and deploy the models safely and responsibly. It says Veo has built-in safety features to help customers block potentially harmful outputs, and encourages best practice on fairness, interpretability, privacy and safety.

Safety filters

The section on safety filters explains the mechanism. Veo accepts prompts in several forms, including text, video and images, and the prompts are evaluated against a list of safety filters that include harmful categories, with violence, sexual, derogatory and toxic as examples. The documentation states the purpose of the filters: to filter input images and videos that violate Google Cloud’s Acceptable Use Policy, the Generative AI Prohibited Use Policy or Google’s AI Principles.

It tells developers what a trip looks like:

  • the model may respond with an error such as the prompt could not be submitted, or that it might violate policies, which means the input triggered a safety filter;
  • if fewer videos than requested are returned, some generated outputs were blocked for not meeting the safety requirements;
  • the output may contain a support code, and each code corresponds to a specific harmful category.

The categories

A table maps support codes to categories. The categories, with the page’s descriptions, are:

Category Description
Child Rejects requests to generate content depicting children unless a person generation setting allows it or the project is allowlisted.
Celebrity Rejects requests for a photorealistic representation of a prominent person unless the project is allowlisted.
Video safety violation General safety violation.
Dangerous content Potentially dangerous content.
Hate Hate-related content.
Prohibited content Prohibited content related to child safety or other sensitive content.
Third-party content Protections related to third-party content.
Sexual content Sexual or suggestive content.
Toxic Toxic content.
Violence Violent content.
Vulgar Vulgar content.
Other Miscellaneous safety problems with the request.

Two of these matter most to adult game work. The sexual content category is defined as “sexual or suggestive”, which goes further than the explicit content in the Prohibited Use Policy. And the child category blocks depictions of children by default, which is a separate hurdle for any story that includes young characters.

What the Prohibited Use Policy says

The Generative AI Prohibited Use Policy shows a last modified date of 17 December 2024. In its third group of rules, it tells users not to engage in sexually explicit, violent, hateful or harmful activities, and the list of content includes sexually explicit content, for example content created for the purpose of pornography or sexual gratification. Other groups bar facilitating non-consensual intimate imagery, content relating to child sexual exploitation or abuse, and circumventing abuse protections or safety filters, “such as manipulating the model to violate our policies”.

It closes with a flexibility sentence: Google may make exceptions based on educational, documentary, scientific or artistic considerations, or where harms are outweighed by substantial benefits to the public.

The same policy lists misrepresenting the provenance of generated content as created solely by a human, with intent to deceive, as prohibited. For game studios publishing AI-made trailers, that links to store disclosure rules.

The Gemini app guidelines

The policy guidelines for the Gemini app describe Google’s aims for the app’s output. They say Gemini should not generate outputs that describe or depict explicit or graphic sexual acts, sexual violence or sexual organs in an explicit manner, with explicit adult material and depictions of sexual assault as examples, and that context matters, including educational, documentary, artistic or scientific applications. The page read shows no date. These guidelines are for the Gemini app. The pages read do not say whether they govern Veo in other products.

How the documents fit together

Three layers appear in the pages read.

  1. The policy. The Prohibited Use Policy sets what users must not do with Google’s generative AI.
  2. The product filters. The Veo documentation describes a technical layer that screens prompts and blocks outputs by category.
  3. The app guidelines. The Gemini page describes what the Gemini app should and should not produce.

The pieces are consistent on explicit content. They differ in wording on suggestive content: the Veo filter category names it, the policy example does not, and the Gemini page targets explicit and graphic content. The documents do not reconcile these.

What this means for adult game developers

  1. Do not plan to make explicit video with Veo. The policy and the filter category both bar it.
  2. Expect suggestive prompts to be blocked. The filter category is “sexual or suggestive”.
  3. Do not try to engineer around a refusal. Circumventing safety filters is its own violation.
  4. Expect problems with young characters. The child category blocks depictions of children by default.
  5. Avoid real people. The celebrity category blocks photorealistic prominent people.
  6. Disclose AI video where stores ask. See AI-generated content rules for adult games, does itch.io allow AI-generated games and, for video hosting, does YouTube allow adult game trailers.
  7. Compare other tools. See does Gemini allow NSFW writing, does Midjourney allow NSFW images and does Adobe Firefly allow NSFW content.
  8. Make the explicit edition without it. Then use a store that accepts it. See where to sell adult games.

What is still unknown

  • How strict the “suggestive” filter is in practice. The page defines the category and gives no examples.
  • Whether consumer products have different rules. The pages for Flow and the Gemini app’s video features were not read.
  • How the Gemini API terms treat Veo. They were not read.
  • What penalties apply. The pages read describe filters, not account action.
  • How the artistic exception is applied. The policy names it without a test.

How this page was put together

The Google Cloud Veo page was read in a browser on 7 November 2026, and the page shows a last updated date of 7 November 2026. It was served in machine-translated Portuguese, and the table and descriptions above are a translation, so short phrases in quotation marks should be read as translations. The Prohibited Use Policy and the Gemini app guidelines were read on 31 October 2026, also in Portuguese, and are described in the Journal’s page on Gemini and NSFW writing.

The author has not generated video with Veo, and nothing here is a first-hand account of its filters. Google Cloud’s Acceptable Use Policy, the Gemini API terms and the consumer terms for Flow were not read.

This page was written on 7 November 2026 and will be updated when Google changes its pages. To report an error, see the corrections standard.

Questions

Does Google Veo allow NSFW video?

Not sexually explicit video. Google's Generative AI Prohibited Use Policy bars sexually explicit content, such as content created for pornography or sexual gratification, and Veo's safety filters include a sexual category described as sexual or suggestive content.

What happens when a Veo prompt trips a safety filter?

The Google Cloud documentation says the model may respond with an error such as the prompt could not be submitted or that it might violate policies. If fewer videos than requested come back, some outputs were blocked for not meeting safety requirements.

What are Veo's safety filter categories?

The documentation lists categories including child, celebrity, dangerous content, hate, prohibited content, third-party content, sexual content, toxic, violence and vulgar, with support codes mapped to each. Prompts are evaluated against filters that include violence, sexual, derogatory and toxic categories.

Can Veo generate videos of children or celebrities?

The documentation says requests to generate content depicting children are rejected unless a setting allows it or the project is allowlisted, and requests for a photorealistic representation of a prominent person are rejected unless the project is allowlisted.

Do the same rules apply in consumer apps such as Flow or Gemini?

The pages read cover the Google Cloud product. The Prohibited Use Policy applies to Google products that refer to it, and the Gemini app has its own policy guidelines. Consumer Veo terms were not read for this page.

Is there an exception for artistic work?

The Prohibited Use Policy says Google may make exceptions based on educational, documentary, scientific or artistic considerations, or where harms are outweighed by substantial public benefits. The Veo page does not describe an exception.

Can I trick Veo's filters?

The Prohibited Use Policy bars generating or distributing content that facilitates evading protections against abuse or safety filters, including manipulating the model to violate Google's policies.

Sources

  1. Google Cloud: Responsible AI for Veo video generation (page last updated 7 November 2026)
  2. Google: Generative AI Prohibited Use Policy (last modified 17 December 2024)
  3. Gemini: Policy guidelines for the Gemini app