OpenAI deploys OpenAI digital watermark technology under EU's new transparency rules

OpenAI is deploying OpenAI digital watermarking technology under the EU's new transparency rules. A new era for artificial intelligence content detection.

OpenAI is preparing to introduce OpenAI digital watermark technology to comply with the European Union's new transparency rules. This step aims to make the texts and codes created by generative artificial intelligence models identifiable.

The company announced that it will benefit from a special system called textGrain in this process. This technology helps identify the source of content by adding invisible statistical signals to models' word choices.

textGrain Technology and Working Principle

The textGrain system developed by OpenAI stands out as a method that enables the detection of content created by artificial intelligence. The system proves that this content is the product of artificial intelligence by placing an invisible statistical signal on the model's word choices.

This signal can be scanned through a special detector and it can be determined whether the content is watermarked or not. OpenAI states that this system matches or exceeds the performance of other approaches such as SynthID.

However, the company warns that strong performance in ideal conditions does not always guarantee reliable results in daily use. Especially short texts and mathematical contents are more difficult to detect.

According to OpenAI data, the success rate for this type of content remains at approximately 80 percent. Edits made to the text are also among the factors that reduce the success of detecting the watermark.

EU Artificial Intelligence Law and Implementation Process

Article 50 of the European Union's Artificial Intelligence Act requires providers of generative AI to make their text output identifiable in a machine-readable way. This regulation covers all existing organizations as well as new companies entering the market.

Companies such as OpenAI and Anthropic, as well as giants such as Microsoft, Google and Meta, must also comply with these rules. The final compliance date for existing organizations was announced as December 2.

The watermark feature will be enabled by default for eligible ChatGPT and Codex outputs in the European Union. Users in other regions will be able to optionally choose this feature for certain models.

OpenAI initially took a more cautious approach to accessing its detection software. Use of the software will initially be limited to approved researchers and expert organizations only.

The company emphasizes that this detector does not identify the user or reveal any prompts or conversations. In addition, it is planned to make the developed technology open source and allow others to work on it.

OpenAI has not yet shared clear details about which models will be available for general use. This technological development is considered an important milestone in the transparency of artificial intelligence content.

Do you think such invisible watermarks added to content produced by artificial intelligence will be enough to ensure security in the digital world?