OpenAI has developed a tool to detect ChatGPT content with 99.9% accuracy

When OpenAI launched ChatGPT in late 2022, teachers expressed concern that students would use the platform to cheat on assignments and tests. To prevent this, numerous companies have deployed AI-generated content detection tools, but they provide reliable results.

Now, OpenAI has announced that it has developed a method for detecting the use of ChatGPT for writing content. The technology is claimed to be 99.9% effective and essentially uses a system that can predict which word or phrase (called a «token») will come next in a sentence. The AI detection tool slightly alters the tokens, which leave a watermark. This watermark cannot be recognized by the human eye, but it can be recognized by a suitable tool.

This technology was ready to be deployed about a year ago, but the company has not yet done so because the project was met with mixed reactions internally. On the one hand, the launch of this tool could alienate part of ChatGPT’s user base. At the same time, the introduction of the tool will mean that the company is fulfilling its commitment to transparency.

At the same time, a representative of OpenAI reported concerns that the tool may «disproportionately affect groups for whom English is not their first language». Nevertheless, key OpenAI staff members who support the launch of the tool believe that the technology can bring many benefits and should not be delayed.

Source: androidauthority