Beware of Copying ChatGPT: The EU Plans to Set a Trap First
Artificial intelligence (AI) makes it easier to create a range of content, from images and audio to writing that resembles human work.
AI-generated visual content, such as images, videos and audio, may be easy to identify. But AI-generated text is more difficult to detect.
Before long, AI-generated text may leave a “trace”. ChatGPT’s parent company, OpenAI, is preparing a new technology that will allow AI-generated text to be detected using a dedicated system.
According to Channel News Asia (CNA), OpenAI has introduced watermarking for text generated by ChatGPT and Codex. The technology will be launched first in Europe in response to the EU AI Act, which requires AI providers to make AI-generated text identifiable in a machine-readable format.
Over the coming weeks, OpenAI will implement invisible watermarks in eligible ChatGPT and Codex text outputs in the European Union. API customers worldwide will later have the option of enabling text watermarking for certain models.
The AI start-up has also begun accepting applications for access to its text-watermark detector. Initially, access will be limited to approved researchers and expert organisations to help test the technology.
OpenAI said its watermarking technology, called “textGrain”, adds a statistical signal to the model’s choice of words. The detector searches for that signal to verify whether text was generated using ChatGPT or Codex.
OpenAI says textGrain will match or outperform all other approaches it has tested. However, the company acknowledges that the technology will not always work as expected.
“The detector can make two types of errors: it can report a watermark when there is none, known as a false positive, or fail to detect a watermark that is present, known as a false negative,” OpenAI said on its website, as quoted by CNA on Saturday (10 October 2026).
OpenAI also acknowledges that shorter texts are harder to detect and that editing can weaken the watermark.
In addition, a watermark detector cannot determine how much a human contributed to a text, who owns it or created it, or whether the information it contains is accurate.
For now, OpenAI plans to introduce text watermarking to all users across every ChatGPT plan in Europe. It has no plans to launch the system globally yet.
OpenAI said the regional approach will give it room to learn from feedback gathered through real-world use of the technology.
Source: jawawa.id (https://jawawa.id/newsitem/beware-of-copying-chatgpt-the-eu-plans-to-set-a-trap-first-1791629430)
Originally published by CNBC Indonesia (https://www.cnbcindonesia.com/tech/20261010114133-37-774931/hati-hati-contek-chatgpt-uni-eropa-mau-pasang-jebakan-batman-duluan), translated from Indonesian