ZICQ
中 Log in / Sign up
Newsroom LLMs & Foundation Models #OpenAI #ChatGPT #Codex #Text Watermarking #AI Compliance

OpenAI Introduces Text Watermarking for ChatGPT and Codex to Meet EU AI Act Compliance

Avatar of Mr.Xu

By Mr.Xu

Published:

中文阅读 (Chinese) English Version

Summary:OpenAI has announced the rollout of its textGrain watermarking technology for ChatGPT and Codex, aimed at enhancing traceability of AI-generated content. Initially available to users in the European Union, this move aligns with the EU AI Act's compliance requirements. The textGrain watermark is described as 'invisible and machine-readable' and is reported to match or exceed the performance of similar technologies like Google DeepMind's SynthID. While OpenAI asserts that the watermark does not co


Key Content

OpenAI has announced the introduction of a new text watermarking technology called textGrain for its AI models, ChatGPT and Codex. The primary goal of this technology is to enhance the traceability of AI-generated text, enabling users and regulators to identify AI-generated content and mitigate potential misuse.

Technical Highlights

  1. Invisible and Machine-Readable: The textGrain watermark is embedded in the text in an invisible manner and can only be detected through specific machine detection methods. This ensures that the watermark does not interfere with the normal reading and usage experience of the text.
  2. Superior Performance: OpenAI claims that textGrain matches or exceeds the performance of similar technologies like Google DeepMind's SynthID, effectively marking AI-generated content.
  3. Compliance-Driven: The introduction of this technology is primarily driven by the need to comply with the EU AI Act, which imposes strict requirements on the transparency and traceability of AI-generated content. The textGrain watermark is a key measure for OpenAI to meet these requirements.

Industry Impact

  1. Regulatory Compliance: As the global AI regulatory environment becomes increasingly stringent, AI companies need to adopt more technical measures to ensure compliance. The introduction of the textGrain watermark sets an important technical benchmark for the AI industry.
  2. User Trust: By enhancing the traceability of AI-generated content, OpenAI aims to increase user trust in AI technology. This is crucial for the long-term development and widespread application of AI.
  3. Technical Challenges: Although the textGrain watermark represents a technical breakthrough, OpenAI acknowledges that it cannot completely prevent the misuse of AI-generated content. This indicates that AI companies still face numerous challenges in both technology development and regulatory response.

Developer Recommendations

For developers, the introduction of the textGrain watermark technology means paying more attention to the compliance and security of AI-generated content. It is recommended that developers:

  • Familiarize themselves with and understand the technical principles and application scenarios of the textGrain watermark.
  • Use the watermark technology in conjunction with ChatGPT and Codex for content traceability and risk assessment.
  • Stay updated on changes in AI regulatory policies and adjust development strategies accordingly to ensure compliance.

Conclusion

The textGrain watermark technology introduced by OpenAI is an important measure for the AI industry to address global regulatory challenges. Although the technology cannot completely eliminate the risk of misuse of AI-generated content, it is significant in terms of improving traceability and enhancing user trust.


Source: The Verge AI (2026-10-05)

— END —

Tags: #OpenAI #ChatGPT #Codex #Text Watermarking #AI Compliance

Community Comments

Loading live comments and annotations…