Live · 7am IST · DailyFeatured
Reel

The ShiftMaker

AI Intelligence Daily
Featured

Claude will begin digitally watermarking AI-generated text and images to comply with the EU's Artificial Intelligence Act

The implementation will include a hidden statistical pattern in text, but image watermarking remains absent. The EU's Artificial Intelligence Act mandates stricter oversight of AI systems, requiring transparency and accountability measures.

Published 12 August 2026 · ID 2026-08-12-claude-will-begin-digitally-watermarking-ai-generated-text-and-images-to-comply-

Anthropic, the company behind Claude, has announced plans to digitally watermark AI-generated text and images as part of its compliance strategy with the EU's Artificial Intelligence Act. This move aligns with the act's requirements for transparency and traceability of AI-generated content. The watermarking process will involve embedding a hidden statistical pattern into text outputs, making it easier to identify content generated by AI systems.

The EU's Artificial Intelligence Act, set to take effect in 2026, imposes stringent regulations on AI systems, particularly those deemed high-risk. These regulations require AI developers to implement measures that ensure the traceability and accountability of AI-generated outputs. Anthropic's approach includes publishing detailed guidance on how Claude will comply with these requirements, ensuring clarity for users and stakeholders.

According to the guidance article published by Anthropic, the watermarking process will be applied to text outputs starting in 2026. The act requires AI systems to be transparent about their use, and the watermarking feature is a critical step in meeting these obligations. While text watermarking is being implemented, image pixel watermarking remains absent, highlighting a gap in the current compliance strategy.

The introduction of digital watermarking may increase the cost of AI development and deployment, as companies must invest in new technologies to meet regulatory requirements. It could also lead to increased vendor lock-in, as users may become dependent on specific AI systems that support watermarking. Additionally, the governance of AI systems may become more complex, requiring ongoing compliance efforts and adjustments to ensure adherence to evolving regulations.

As the EU's Artificial Intelligence Act moves closer to implementation, companies like Anthropic are taking proactive steps to ensure compliance. The watermarking initiative is a significant development in the AI industry, setting a precedent for transparency and accountability. While the focus remains on text outputs, the absence of image watermarking underscores the need for further innovation and regulation in this area.

Sources

Share on X Share on LinkedIn