News Technology

OpenAI Rolls Out Text Watermarking for ChatGPT and Codex in the EU

OpenAI has launched an invisible, machine-readable text watermarking system called textGrain for ChatGPT and Codex, with an initial rollout limited to users in the European Union.

The watermarking technology embeds a detectable signal into AI-generated text that can be identified by specialized tools without altering the visible content. OpenAI claims textGrain "matched or exceeded" other approaches such as Google DeepMind's SynthID for text, which also serves as the foundation for watermarking Anthropic announced in August.

Both OpenAI and Anthropic implemented text watermarking in response to requirements under the EU's AI Act, which establishes transparency obligations for providers of general-purpose AI models. The legislation mandates certain technical measures to distinguish AI-generated content from human-created material.

OpenAI published benchmark data indicating that watermarked text performs comparably to unwatermarked output across standard evaluations. However, the company acknowledges that textGrain does not provide absolute certainty in detecting AI-generated content.

The staged EU-first deployment suggests a broader rollout may follow as other jurisdictions develop similar regulatory frameworks around AI transparency and content attribution.

Sources

The Verge

OpenAI Begins Rolling Out Text Watermarking for AI-Generated Content

OpenAI is moving forward with text watermarking as a compliance measure under the EU AI Act. The feature, which embeds a detectable signal into AI-generated text, aims to help distinguish machine-produced content from human-written material. The watermarking approach addresses the Act's transparency obligations, which require providers of general-purpose

Nobel Prize Awarded for Optogenetics: Controlling the Brain with Light

The Nobel Prize in Physiology has recognized the groundbreaking work in optogenetics, a technique that enables scientists to control brain cells using light. The foundation of this revolutionary method traces back to research on algae that naturally move toward light sources. By introducing light-sensitive proteins from these organisms into neurons,

Agentic AI Transforms Microscopy: Simple Language Commands Power Self-Guided Experiments

Researchers at Argonne National Laboratory are pioneering a new approach to microscopy that integrates agentic AI to interpret natural language commands and autonomously guide experimental processes. Traditional microscopy techniques often require specialized expertise to operate advanced instruments and design experiments. This new AI-driven system aims to democratize access by allowing

Sources

Read next