OpenAI introduces watermarks for ChatGPT text in the EU

OpenAI has announced that it will watermark texts from both ChatGPT and Codex in the European Union. This step is a response to the new AI legislation, the AI Act, which sets requirements for transparency and recognition of AI-generated content. The aim of this measure is to ensure that users can identify whether a text has been created by an AI model or not.

The watermarks are designed to be subtle and difficult to detect, so they do not interfere with the user experience. Nevertheless, OpenAI has warned that editing texts may weaken these watermarks, making it more challenging to identify AI-generated content. This raises questions about the effectiveness of watermarks in ensuring transparency when texts are further modified.

Moreover, OpenAI discusses the obligations under the AI Act, which interested parties in the technology and legal community are actively monitoring. With this measure, OpenAI anticipates growing concerns about the ethics and responsibility of AI technologies. This not only underscores their compliant approach but also their focus on building trust between users and AI.

The watermarking technology is still under development, and users are eager to learn more about how this method will function in practice. It is clear that the implementation of the AI Act and the subsequent measures from companies like OpenAI will significantly influence the way AI content is produced and consumed.

Read the full article from Techcrunch.