SpaceXSI: Grok text API output watermarking on by default from November 7
SpaceXSI docs state that from November 7, 2026, text API responses from several Grok models include an invisible watermark by default to comply with the EU AI Act. No request changes are required; team admins can opt out per model in the Console.
SpaceXSI (xAI) developer docs page “Output Watermarking” states that starting November 7, 2026, text-generation API responses from these models include an invisible watermark by default: grok-4.7, grok-4.6, grok-4.5, grok-4.3, and grok-4.20-reasoning, grok-4.20-non-reasoning, and grok-4.20-multi-agent.
SpaceXSI says it watermarks text model output to comply with the EU AI Act. Watermarking is enabled on every SpaceXSI API endpoint, including the global endpoint and regional endpoints. It applies to text output from the Responses API and the Chat Completions API, covering ordinary text, structured outputs, and tool-call arguments. The watermark is applied while the model generates text by slightly shaping token choices; it adds no characters, markup, or hidden text. SpaceXSI says it does not change response quality or speed.
Integrations need no request-parameter changes. Read the x-watermark response header: true means that response was watermarked. Access to watermark detection is granted by application.
Team admins can turn watermarking off for some or all models in the SpaceXSI Console; changes apply to new requests within five minutes. Opting out requires confirming that an exception under applicable law applies, and that the team is responsible for any marking or disclosure duties. Some models may be locked on for a team.
This is a change scheduled to turn on by default on November 7, not a watermark requirement that is already fully live for all listed traffic.