Announcing MCP Toolbox Java SDK v1.0: Agentic data access for the enterprise
cloud.google.com·6d ago
Cloud & Infrastructure
The New Stack·7d ago·5 min read
OpenAI brings text watermarking to its API — and unlike Anthropic, it’s off by default
TL;DR
OpenAI has introduced an optional text watermarking feature for its API, allowing developers to identify AI-generated content. Unlike Anthropic's approach, this feature is disabled by default, giving users control over its implementation.
✦ Why It Matters
Consider enabling watermarking in your applications to enhance content authenticity and transparency.
Key Takeaways
- Enable text watermarking in your OpenAI API projects for compliance.
- Understand the limitations of watermarking in shorter texts and code.
- Monitor the performance of watermarked outputs in your applications.
- Stay informed about OpenAI's updates on textGrain and its open-source plans.
How It Works
textGrain embeds a statistical signal in generated text by subtly influencing word choices, creating a detectable pattern over longer passages. This allows OpenAI's detector to identify watermarked content effectively.
⚠ The Catch
Detection rates drop significantly when editing occurs, with a 10% word replacement reducing detection from 92% to 66%, and 25% replacement dropping it to 17%.