Anthropic adds SynthID-Text watermarking to Claude

On August 14, 2026, Anthropic disclosed that Claude embeds text watermarks using a version of Google DeepMind's SynthID-Text, first published in Nature in 2024. The watermark subtly alters word-sampling randomness during generation, creating a pattern invisible to readers but detectable with a key. This is among the first production deployments of AI text watermarking in a major consumer assistant.