WRITER adds customer-specific invisible watermarks to Palmyra model text

In a press release dated October 27, 2026, WRITER introduced text watermarking for output from its Palmyra models. When enabled, generated text carries an invisible mark tied to the customer's organization, which a detector can read and decode. WRITER says customers control watermark strength, which content types are marked, and whether custom data such as a campaign ID is embedded. EU customers get watermarking on by default, configured for EU AI Act requirements. Non-EU customers can opt in. CEO May Habib said in the release that enterprises want ownership and configurability over how their AI-generated content is watermarked.

The details

• Applies to text generated by WRITER's Palmyra models
• Watermark is invisible and tied to each customer's private key, per the company's roundup post
• Adjustable strength trades detection reliability against content quality
• Customers can watermark some content types (for example blog posts) and not others (for example ad copy), where regulations allow
• Multi-bit approach can embed custom information such as a campaign ID
• On by default for EU customers; optional for others; available December 2, 2026

Context

The announcement follows WRITER's September 9, 2026 launch of Enterprise Brain, a context layer for its agents, and the August 13 release of the Palmyra X6 model with agent harness upgrades. Earlier in 2026 the company released an agent Skills creator and appointed Brian O'Reilly as its first chief operating officer. WRITER's October roundup notes that when third-party models are used in WRITER, that provider's watermark policy applies.

Not disclosed

• Pricing or plan requirements for watermarking
• Detection accuracy figures at different strength settings
• Whether the detector is available to third parties or only to the customer
• How watermarks hold up after text is edited or paraphrased