OpenAI doesn’t want to add anti-plagiarism tools to ChatGPT’s text, claims insider

Ever since ChatGPT began making waves online, people have been concerned about the power of AI writing when it comes to plagiarism. With AI tools, people don’t need to write their essays and reports; instead, they can plug the subject into an AI generation tool and have it write everything for them. As it turns out, OpenAI has the technology to make it a lot easier to detect when ChatGPT generates a passage of text; however, internal conflict surrounding the tool has prevented the company from actually deploying it.

Image Credit: OpenAI

As reported by the Wall Street Journal, insiders claim that OpenAI has the power to add “fingerprinting” to the text that ChatGPT generates. Fingerprinting is when an AI developer adds a tell-tale pattern to the model’s generation which humans can’t detect. Then, the developer can create a tool that looks for this special pattern and flags anything that features it as AI-generated. People concerned about AI use can then use this tool to spot AI generation in ways that a human can’t.

The fingerprinting tool that OpenAI made for ChatGPT reportedly has a 99.9% detection rate, which would help cut down on people using it for plagiarism. There’s just one problem; some employees don’t think it should be deployed. Some cite that AI generation is handy for people who aren’t native English speakers to get their point across, while others point to a recent survey where 1/3rd of OpenAI’s paying members said they do not want fingerprinting technology applied to their generations.

The debate has been raging for two years now, and the company still hasn’t come to a solid conclusion on what it wants to do. Until then, the fight between plagiarisers and the people trying to catch them continues to wage on.