technology

Anthropic details implementation of invisible watermarks for Claude-generated text

17 Ağustos 2026The Verge

Anthropic has detailed its plan for implementing invisible watermarks on text generated by its AI model, Claude, to adhere to European AI transparency regulations. The company revealed that this watermarking system is based on the SynthID-Text approach, an open-source technology developed by Google DeepMind.
This method utilizes detectable patterns formed by analyzing wording probabilities, ensuring that the generated content can be traced back to its source.
As AI technologies proliferate, regulatory bodies, especially in Europe, are pushing for measures that ensure transparency and accountability in AI-generated content. The implementation of watermarking systems is seen as a proactive approach to address concerns regarding misinformation and the ethical use of AI.
The introduction of invisible watermarks in AI-generated text represents a significant step toward accountability in AI technologies. By adopting the SynthID-Text approach, Anthropic not only aligns with regulatory requirements but also enhances the integrity of AI outputs.

NewsAI özeti

This summary is for informational purposes only and does not constitute professional advice.