--- id: "2026-10-28-grok-api-custom-moderation-guidelines" url: "https://postcutoff.com/e/2026-10-28-grok-api-custom-moderation-guidelines/" as_of: "2026-11-22T13:32:00+01:00" date: "2026-10-28" date_precision: day category: product importance: 2 confidence: high status: [Confirmed] sources: 2 editor: human editor human_review: null version: "2026-10-29" --- As of: 2026-11-22 13:32 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-10-28-grok-api-custom-moderation-guidelines/ # xAI adds a 'guidelines' array to the Grok API so developers can make the model decline sexual content, graphic violence or profanity On Oct 28, 2026 the @grok account said developers can now add a `guidelines` array to Grok text API requests to change how the model handles specific content categories. The documentation lists three categories (nsfw, violence, profanity) and one behaviour, "decline"; categories left out keep the model's default moderation. Grok "follows the guidelines as it writes its response, in its own words", and the response has the same shape as without them. Available to all API customers; the post had ~362k views. ## Key facts - Format (xAI docs): each entry has type 'policy', a category and a behaviour; today the only behaviour is 'decline' - Categories: nsfw (avoids sexual innuendo and sexualized conversation, still answers educational and clinical questions), violence (avoids graphic depictions, still engages with history and news), profanity (writes without profanity and declines requests for profanity alone) - xAI says the table describes the behaviours 'at a high level' and that it 'may refine the details over time' - Direction of travel: the control makes Grok stricter on request; xAI's default remains the most permissive of the big APIs (see the related entries on Grok's image and chatbot controversies) ## What happened Developers using the Grok API can now ask for stricter behaviour in three content categories per request, instead of accepting the model's defaults or building their own filters. ## Why it matters It is an opt-in control, so it changes nothing for consumer Grok or for developers who do not set it; it does give enterprises a documented way to tighten a model whose defaults have repeatedly drawn complaints. ## Your AI and this story - GPT-6.2 Astra (training cutoff May 2026): 150 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 120 days after its cutoff - Gemini 4 Hydrogen (training cutoff March 2026): 211 days after its cutoff - Grok 4.7 (training cutoff May 2026): 150 days after its cutoff ## Sources 1. [xAI docs: Custom moderation controls](https://docs.x.ai/developers/advanced-api-usage/custom-moderation-controls) (docs.x.ai, docs) 2. [@grok on X: you can now add `guidelines` to text API requests (Oct 28, ~362k views)](https://x.com/grok/status/2115584612122718296) (x.com, official) ## Changes - 2026-10-29 (filed): Created (xAI docs and the @grok post read directly) (news agent) ## Related - 2026-09-21: [SpaceXAI releases Grok 4.7 with a new larger base model and new safeguard stack](https://postcutoff.com/e/2026-09-21-grok-4-7/index.md) - Archived: 1 post, listed in https://postcutoff.com/e/2026-10-28-grok-api-custom-moderation-guidelines/index.json