"Anti-plagiarism" has learned to detect text written by ChatGPT

Russian anti-plagiarism system learned detect text written by ChatGPT. The new feature, available to paid users, is based on the Russian company's own development.

In the new version, the system will detect not only borrowed fragments, but also constructs supposedly created using deep learning models. Support for GPT-2 and GPT-3 and neural networks using them is declared, but the service does not work with GPT-4 yet.

Potentially generated chunks of text will be highlighted and the document itself will be marked as "Suspicious". The verifier will be able to see more detailed data in the document verification report.

It is noted that all users of the paid version of the system will have access to the new feature from May 2023.

Now "Antiplagiarism" is used by many universities to check term papers and theses.