Autonomous AI agents generate hundreds of scientific preprints in twelve weeks
Researchers tested 936 autonomous AI agents across 35 laboratories, generating hundreds of scientific preprints and raising serious concerns about academic peer review and paper verification.
FTMQ Science, written by our newsroom1 view
On November 19, 2026, researchers from Stanford University, MIT, ETH Zurich, and alphaXiv unveiled the results of a 12-week study involving autonomous artificial intelligence agents, CASRAI reported. An autonomous agent is an artificial intelligence system capable of performing complex tasks independently. According to CASRAI and eu.36kr.com, the project deployed 936 intelligent agents that produced 936 scientific preprints. [1][2][3]
The experiment spanned 35 independent research laboratories, CASRAI noted. In addition to producing papers, the autonomous agents established an automated peer-review journal. Eu.36kr.com reported that the platform features a built-in peer-review mechanism designed to process the large volume of generated manuscripts. [1][2]
Automated scientific generation previously made headlines on October 6, 2026, when OpenAI released 722 manuscripts claiming solutions to open mathematical problems via a GitHub repository. That event raised widespread concerns among mathematicians regarding validity and mathematical exposition style. [4]
In the latest study, the machine-generated research achieved high computational execution, CASRAI reported. However, verifying the findings of autonomous agents requires audits of executable code. CASRAI noted that this requirement introduces fundamental governance challenges for preprint repositories, university research offices, and academic review integrity. [1]
In short
- Researchers deployed 936 autonomous AI agents across 35 independent research laboratories over a 12-week period.
- The AI agents produced 936 scientific preprints and established an automated peer-review journal.
- Verifying the findings generated by the autonomous agents requires executable code audits.
Sources
Every paragraph above points to the numbered items it rests on. Read the originals here.
- [1]Autonomous Agent Platform alphaXiv Produces 936 AI-Written Preprints in 12-Week StudyCASRAI, 21h ago (the report this story comes from)
- [2]Agent Edition arXiv Launched: 936 Intelligent Agents Generate Nearly 1000 Papers in 12 Weeks with Built-In Peer Review Mechanismeu.36kr.com, 1d ago
Background
- [3]Autonomous agent on Wikipedia
- [4]OpenAI bulk release of mathematical papers on Wikipedia
- [5]Week on Wikipedia
Our newsroom writes these reports with the help of software, from the 5 sources listed and nothing else, and checks them against those sources. Facts can still be wrong or move on; the originals are the record. Spotted a mistake? Write to daniel@monsterkong.com.
Related from FTMQ Science
Earlier reports of ours on the same people and subjects.
- Zapata Quantum Launches ForShor AI Verification Challenge for Quantum AlgorithmsTop Stories, 2d ago
- Researchers present framework for evaluating economic impact of European Big ScienceTop Stories, 13d ago
- Researchers develop artificial intelligence model to accelerate laser interactions simulationPhysics, 6d ago
- AI education research increased tenfold over five years in four journalsNew Research, 6d ago
- Perseverance rover uses artificial intelligence software to schedule tasks on MarsEarth Science, 22d ago
More in New Research
- Researchers identify over two thousand indoor arthropod species across North America15h ago
- Astronomers observe giant star swallowing stellar companion in Andromeda Galaxy15h ago
- Handheld Raman spectroscopy tested for ex vivo cervical lesion assessment1d ago
- Male mice gain defensive benefits from observing aggressive encounters1d ago
- Large language models show accuracy failures when prompted on climate information1d ago
- Brain processes food rewards similarly to monetary gains study finds1d ago
Get the day in one email
Reports like this one, the top news of the last 24 hours, every morning. Free, one email a day; readers can comment under every report.
By signing up you agree to our terms and privacy policy. Unsubscribe any time.

Comments
Loading
Join the conversation
Comments are open to readers of our daily email: the top news of the last 24 hours, every morning, free. Sign up and the comment box opens.
Already on the list? Enter the same address and we will send a sign-in link.
By signing up you agree to our terms and privacy policy. Unsubscribe any time.