News for Agent

Daily business news: funding, M&A, launches

Product Launches

xAI Turns On Invisible Text Watermarking for Grok API Models by Default

By ·

Key takeaways

  • xAI will enable invisible statistical watermarking across its Grok API models by default starting November 7, 2026.
  • The watermark covers grok-4.7, grok-4.6, grok-4.5, grok-4.3 and grok-4.20 and every text generation endpoint.
  • Team administrators can disable the watermark per model through the xAI Console for legal exceptions and compliance reasons.

xAI will switch on invisible statistical watermarking across its Grok API models by default starting November 7, 2026, the company said in an announcement shared on X. Crypto Briefing reported that developers will not need to change a setting. The move leaves ChatGPT as the only model in North America that does not watermark its text, according to the announcement.

How the Grok Watermark Covers the API

The feature applies to a list of model versions: grok-4.7, grok-4.6, grok-4.5, grok-4.3 and grok-4.20. Coverage extends to every text generation endpoint, including the main api.x.ai endpoint, the Responses API and the Chat Completions API.

The watermark works by nudging token selection toward a pattern as the model picks its next word. A single sentence looks unchanged, but running enough text through a detector surfaces a statistical fingerprint. xAI says the process does not change the length, quality or speed of responses.

The mark is the default but not mandatory for everyone. Team administrators can disable it on a per-model basis through the xAI Console, an opt-out framed for specific legal exceptions and compliance reasons rather than a general off switch. Consumer-facing products are not affected; the change targets the API, where businesses and developers plug Grok into their own apps and workflows.

Before this change, xAI had publicly pushed back on watermarking outputs, arguing that text should remain free of alterations. The new approach keeps the visible output unaltered while the fingerprint lives in the statistics. The EU AI Act requires greater transparency from generative AI systems, and the move puts xAI alongside peers such as Anthropic in meeting those obligations.

Grok 4.5 and the Developer Partnerships Behind the API

According to akta.pro company records, xAI released Grok 4.5 in July 2026, a flagship reasoning model with a 500k token context window, embedding it in GitHub Copilot and rolling it out across grok.com, X, iOS and Android within two weeks. The same records list a July 2026 partnership with GitHub (Microsoft) that made Grok 4.5 available in GitHub Copilot for millions of developers using VSCode and GitHub products, with model picker integration enabling direct access inside the Copilot workflow. A June 2026 partnership made Grok models available via Amazon Bedrock, giving AWS customers access through existing AWS infrastructure and commitments.

Earlier xAI Moves on Compute and Model Access

On October 1, 2026, Crypto Briefing reported that xAI is renting surplus GPUs to Anthropic, Google and an unnamed client, with Colossus 1 at 11% utilization. On October 21, 2026, EIN Presswire reported that Viirtue added Google Gemini and xAI Grok connectors to its white-label MCP server, letting businesses query phone systems, call centers, AI voice agents and fax, with setup taking about five minutes at no extra cost. On October 28, 2026, Crypto Briefing reported that xAI subsidiary MZX Tech is offering to buy homes in Southaven, Mississippi, contingent on sellers releasing claims in a noise lawsuit, targeting 418 homes with at least five sold since October 2026.

Where Developers Can Find the Grok API

xAI lists its models and API access on its own site. The watermarking change takes effect for Grok API output on November 7, 2026.

Companies in this story

Sources

Machine-readable: Markdown · JSON