Midjourney temporarily bans some words about the human reproductive system to prevent generating shocking or gory images, while the company “improves things”
Midjourney says it's a temporary measure to stop people from using its system to create shocking or gory images.
Context & Ripple Effects
In February 2023 Midjourney reached for the bluntest tool in the moderation drawer: banning words about the human reproductive system outright, framed as temporary while the company "improves things." The move landed amid an industry-wide scramble over explicit generated imagery — weeks later Imgur was preparing its own ban on explicit images, mixing automatic and human moderation.
The pattern this story starts kept repeating at larger scale: by January 2026 xAI had faced enough outcry over sexually explicit and violent Grok imagery that it restricted generation and editing to paid subscribers and blocked edits of real people in revealing clothing, geoblocked where illegal. Midjourney's keyword filter was an early, crude instance of what has since become standard crisis response for image generators.
First-order effects
- Users immediately lose the ability to prompt Midjourney with blocked reproductive-system vocabulary, even for legitimate medical or artistic uses, until the company finishes whatever fix "improving things" produces.
- Midjourney trades precision for speed: a keyword blacklist catches abusive prompts but pushes all edge-case anatomy requests through the same filter, raising false-positive complaints from its creative user base.
Second-order effects
- Competing platforms inherit the same abuse pressure with no shared playbook — Imgur's decision to allow artistic nudity while expecting moderation problems shows each provider calibrating the explicit-content line independently.
- When keyword filtering proves insufficient, the next lever is access itself: Grok's eventual shift of image tools behind a paid tier shows companies converting moderation failures into subscription gating rather than better classifiers.
Third-order effects
- If the pattern holds, generative-image platforms converge on a layered control stack — prompt filters, paywalls, and geographic blocks — where the strictness a user faces depends on what they pay and where they live, not just what they type.
- Repeated moderation crises across providers strengthen the case for external rules over ad-hoc platform policy, feeding regulatory interest in how synthetic imagery of bodies is produced and distributed.
The trend: Generative-image platforms are cycling through increasingly restrictive access controls — keyword bans first, then paywalling and geoblocking — as reactive substitutes for reliable content moderation.