Archive — US — techcrunch
Anthropic says it will watermark text generated by its AI models
Anthropic will implement watermarking for text generated by its AI models to comply with the EU AI Act's Transparency Code, effective August 2. This watermark will ensure AI-created content is identifiable and will apply to all models released after this date, extending to older ones. Other tech companies are following suit to enhance transparency and avoid regulatory issues.
Key points
- Anthropic confirmed watermarking for AI-generated text models including Claude, to comply with EU regulations.
- The EU AI Act's Transparency Code, effective August 2, mandates marking AI-generated content.
- Watermarking applies to all models released after this date and will be automatically embedded.
- Watermarks will persist when text is copied and pasted, potentially surviving some editing.
- The functionality will be available on platforms like Claude API and Claude Code.
- Other AI platforms, including Substack and Suno, are also adopting similar watermarking measures.
Why it matters
- The move is part of a broader trend in the tech industry toward transparency and accountability in AI as regulations tighten.
- Historically, tech companies have faced backlash for the misuse of AI technologies, prompting a shift in operational practices.
- By 2025, AI regulations are expected to further evolve, influencing development strategies in various sectors, including media and entertainment.
In response to regulatory pressures, Anthropic's watermarking initiative illustrates a critical shift towards responsible AI practices, reflecting wider industry movements to ensure transparency and compliance.