IEEE Spectrum reports on AI slop challenges to scientific publishing

Despite the ease of creation, IEEE’s current defenses, which combine human peer review with automated research-integrity checkers, have so far managed to detect and reject these submissions. However, the article notes an emerging concern: as large language models become more advanced, they are making fewer obvious errors, potentially outpacing current detection systems. This creates an arms race between increasingly convincing AI-generated content and the tools designed to identify it.

The issue extends beyond academic publishing to platforms like Wikipedia, where volunteer groups such as WikiProject AI Cleanup are struggling to maintain the encyclopedia's status as a repository of human-generated information. With fewer resources than commercial publishers, these volunteers face significant challenges in filtering out machine-generated text. The article suggests that if trust in these foundational knowledge institutions erodes, it could undermine public confidence in science and shared reality, leaving humans potentially excluded from the loop as AI systems begin communicating with one another.