SpaceXSI API enables text watermarking by default on Grok models Nov 7
wunderwuzzi23 · x · 2026-11-01
Security researcher Johann Rehberger shared an official notice: the SpaceXSI API is adding text watermarking, enabled by default for all teams on November 7, 2026.
Key points:
- Watermarked-by-default models include grok-4.7, grok-4.6, grok-4.5, grok-4.3, grok-4.20-reasoning, grok-4.20-non-reasoning, and grok-4.20-multi-agent.
- Users who believe a legal exception applies can opt out per model via the SpaceXSI Console; everyone else needs no action.
- API text outputs will carry detectable watermarks by default, directly affecting developers building on Grok outputs.
Related event: xAI's Grok API to Enable Text Watermarks by Default(3 posts)→
More from Models
- Google Gemini Quietly Stopped Adding UTM Parameters to Cited URLs — OnlineInference · 2026-11-01
- Unverified buzz: 'Argon' near Opus-tier, 'Terra 6.1' a cheap fast grunt-work model — banteg · 2026-11-01
- SenseNova-Loop-9B-Preview, a looped-transformer multimodal model, trends on Hugging Face — sensenova · 2026-11-01
- Yacine on new models: 'It's good' — using them for Blender lab planning and family tech support — yacineMTB · 2026-11-01
- NVIDIA ships an 81GB FP4 rewrite of DeepSeek-V4.1-Flash with near-identical coding scores — evisoft · 2026-11-01
- Grok now watermarks generated text — users weigh how to avoid it — walkingriver · 2026-11-01