ChatGPT and other AI systems are being used to generate low-quality spam content that is overwhelming magazines, publishers, and potentially polluting the internet with AI-generated material that could degrade future AI training.
The report describes how ChatGPT and other AI language models are being used to generate large volumes of low-quality content that is being submitted to magazines and publishers as spam. Jennifer Stevens, executive editor of International Living magazine, reports filtering out numerous useless AI-generated article pitches. NewsGuard found the number of fake news websites using AI to generate content grew from 49 in early May to 277 by the end of June 2023. These sites appear designed to make money through Google's advertising network. YouTube videos are promoting schemes to use ChatGPT for creating ebooks and blogs filled with AI-generated content to generate ad revenue. Publishers using the Moksha content submission system report an uptick in AI-generated submissions, with tools developed to block such content. The science-fiction magazine Clarkesworld temporarily stopped accepting online submissions due to being overwhelmed by hundreds of AI-generated stories. Publishers describe the AI content as having perfect grammar but being formulaic, bland, and incoherent. Researchers warn this trend could lead to 'model collapse' where AI systems become less useful as they train on increasing amounts of AI-generated content rather than human-created material.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Highly personalized AI-generated misinformation creating “filter bubbles” where individuals only see what matches their existing beliefs, undermining shared reality, weakening social cohesion and political processes.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed