OpenAI adds text watermark for EU ChatGPT users to meet regulations
Summary
OpenAI is set to introduce a text watermark for users of ChatGPT in the European Union, in compliance with EU regulations aimed at ensuring text provenance. This move aligns with the EU AI Act Article 50, which mandates generative AI providers to mark synthetic text outputs in a detectable format. Additionally, OpenAI has committed to the European Commission's Code of Practice on Transparency of AI-generated Content, joining other leading AI providers in implementing multi-layered provenance strategies to enhance transparency in AI-generated content.
Analysis
OpenAI: OpenAI develops generative AI models and consumer services including the ChatGPT platform. The company has built and deployed provenance signals such as C2PA metadata and SynthID watermarks for images and audio outputs. It recently confirmed plans to extend text watermarking specifically for EU ChatGPT users to address regional transparency requirements. Compliance: OpenAI signed the European Commission's Code of Practice on Transparency of AI-generated Content to support marking and detection standards. Regulation: The EU AI Act Article 50 requires providers of generative AI systems to mark synthetic text outputs in a machine-readable format detectable as artificially generated. Industry Trend: Leading AI providers are adopting multi-layered provenance approaches including statistical watermarks for text to meet emerging transparency obligations.
Categories
aitech
Related sources
- https://www.layer3labs.io/guides/does-chatgpt-watermark-text
- https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content
- https://www.lexology.com/library/detail.aspx?g=69722e64-d0ac-404a-a347-f076f92e5003
- https://openai.com/index/supporting-eu-trustworthy-ai-ecosystem/
- https://theairankings.com/guides/what-is-ai-watermarking/
- https://news.lavx.hu/article/openai-backs-the-eu-s-ai-transparency-code-but-the-underlying-provenance-tech-is-still-fragile
- https://help.openai.com/resources/translation-view-redirect?articleId=8912793&routeLanguage=ru-ru&redirectTo=%2Fen%2Farticles%2F8912793-c2pa-and-synthid-in-openai-generated-images
- https://writehuman.ai/blog/does-chatgpt-watermark-text
- https://vorplabs.com/ai-regulatory-updates/ai-content-marking
- https://www.springlex.eu/en/packages/ai-act/ai-act-regulation/article-50/
- https://openai.com/index/advancing-content-provenance/
- https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act
- https://ec.europa.eu/newsroom/dae/redirection/document/130913
- https://aidetector360.com/blog/does-chatgpt-watermark-text