Ukrainian Roman Oliynyk has developed a system that could change the way we check text using artificial intelligence. Unlike conventional AI detectors, his technology doesn’t try to guess whether a text was written by a human or a machine. It creates a kind of digital "passport" for the document.
Ukrainian Roman Oliynyk has developed a system that could change the way we check text using artificial intelligence. Unlike conventional AI detectors, his technology doesn’t try to guess whether a text was written by a human or a machine. It creates a kind of digital "passport" for the document.
According to the patent application, as text is written, the system creates an encrypted digital trail. The text itself is not recorded in this trail. Later, a person with the appropriate key can check whether the document remains original, whether it has been altered, and determine the place of interference. It is possible to verify the authenticity of a document even without revealing its contents, which is potentially important for courts, universities, media and companies.
This is not an AI detector or a technology that “proves who wrote the text.” It proves something else: that the text is the same as the model wrote, and where exactly it was changed. The system does not judge text without a “passport,” and the absence of a passport does not mean that it was written by a human.
In 2026, the problem of detectors became especially noticeable. They have repeatedly wrongly accused people of using AI, and researchers have shown that even a small rephrase makes the text practically "invisible" to known recognition systems. Roman Oliynyk compares his development to a digital signature: "The system does not say that the text is "similar to AI", but makes it possible to verify its origin."
Public access to the service will be opened in November.
In September 2026, the U.S. Patent and Trademark Office accepted Oliynyk’s application under the Track One expedited procedure. The application lists him as the first inventor, and co-authors include Serhiy Nikolaychuk, founder of venture capital studio The Capital Index. Roman is originally from Dnipro, has worked in network security for over a decade, and heads PayCore Media, Inc.
You can’t check arbitrary text: the system doesn’t do that, and that’s the point. Instead, you can ask the model to write the text and try to trick the system: change a word, rearrange a sentence, copy from a chat. Access for the editors will be opened together with the public release in November.
If the technology passes the patent procedure and finds practical application, it could become an alternative to the current approach to combating AI texts: not guessing the author, but giving the document itself the opportunity to prove its authenticity.
