Grok Watermarks Text by Default
Grok’s API will add invisible statistical watermarks to text outputs from supported models starting November 7, 2026, including structured outputs and tool-call arguments. Team administrators can opt out per model through the developer console.
This is a meaningful provenance feature for developers, but it remains a probabilistic signal rather than durable proof of authorship.
- –Applies across Grok’s Responses and Chat Completions APIs.
- –The `x-watermark` response header lets developers verify whether a request was watermarked.
- –Watermarking slightly shapes token selection without adding visible characters or metadata.
- –Detection is statistical and becomes stronger with longer text, making heavily edited outputs harder to classify.
- –The team-level opt-out gives API operators control, but individual users may have little say inside managed deployments.
DISCOVERED
1h ago
2026-10-31
PUBLISHED
1h ago
2026-10-31
RELEVANCE
AUTHOR
mark_k