ChatGPT creator OpenAI has developed internal tools that can watermark and track AI-generated content with 99.9 percent accuracy, but is refusing to release it. 

OpenAI has been holding onto several tools that can flag any piece of work generated with AI. And according to a spokesperson from the company, all they need to do is press a button. 

Per reports by the WSJ, they began discussing the need for a watermarking tool that lets recipients know if a piece of content was created with content, when they released ChatGPT in 2022. The software was created shortly after and can function accurately when applied to enough texts written with ChatGPT. 

“It’s just a matter of pressing a button,” a source from the project told the WSJ.

Public Response To Open-AI Creating An AI Writing Detection Tool

In April 2023, OpenAI conducted a global survey that received massive support amongst the general public for implementing watermarking tools. 

However, another survey conducted on OpenAI users in the same month, revealed that 30% of its customer base said they would boycott ChatGPT if it started adding watermarks for a competitor that didn’t. This second survey was more influential in the company pausing further discussions for the watermarking tool. 

A spokesperson for OpenAI when speaking to the WSJ said;

“The text watermarking method we’re developing is technically promising but has important risks we’re weighing while we research alternatives,” said the spokesperson. “We believe the deliberate approach we’ve taken is necessary given the complexities involved and its likely impact on the broader ecosystem beyond OpenAI.” (After the WSJ’s report was published, TechCrunch noticed that OpenAI updated a previously published blog post about its watermarking efforts.)”

What An AI Writing Watermarking Tool Mean For Institutions

Effective tools for flagging texts generated with AI could be resourceful in many ways for educational institutions. Such tools can crack down on cheating students who use ChatGPT to write essays and also sort through all that AI-generated sludge flooding the web. 

Although an AI-detection tool can prove very useful for educators who are concerned that tools like ChatGPT are taking over students’ reasoning abilities, OpenAI has waffled on whether or not it should release the tool.