Autonomous research system reaches journal submission, shifting the burden to publisher checks
A journal caught two synthetic papers, but the experiment exposed a scalable new burden on peer review.
IEEE Spectrum gave FARS a research topic; the agentic system proposed and ran a computational experiment, then produced a submission-ready paper after a single command to write it. IEEE Access rejected the paper before peer review because its real references leaned heavily on non-peer-reviewed preprints, while a separate chatbot-fabricated paper failed checks for invented citations. The test shows autonomous research can multiply superficially credible submissions, moving scarce scholarly-publishing capacity from evaluating discoveries toward verifying whether the work and evidence can be trusted.