Midjourney temporarily bans some words about the human reproductive system to prevent generating shocking or gory images, while the company “improves things”
Melissa Heikkilä / MIT Technology Review :
Context & Ripple Effects
Midjourney's keyword ban is an early, crude entry in a moderation arc that keeps repeating across generative platforms. The same vocabulary-level problem later surfaced at Meta, where the Oversight Board pushed the company to relabel 'derogatory' terms around explicit AI-generated images as 'non-consensual' — a sign that word choice itself became the battleground.
First-order effects
- Users lose access to prompts touching the reproductive system entirely — medical, educational, and artistic requests are blocked alongside the shocking or gory output the ban targets.
Second-order effects
- Rival platforms face the same pressure and converge on their own fixes: X restricted Grok's editing of real people in revealing clothing and geoblocked it where illegal (Grok's image-editing crackdown), then formalized labeled consensual NSFW under its adult content policy rather than banning outright.
Third-order effects
- The industry appears locked in a restrict-then-formalize cycle — echoing Instagram's 2019 sensitivity screens after self-harm deaths — where blunt keyword bans buy time until platforms settle on labeling, consent framing, and access gating instead of prohibition.
The trend: Generative-image platforms are cycling through blunt vocabulary bans toward durable policy regimes built on consent labels and access tiers as explicit-content misuse forces repeated intervention.