Google Paper: Autonomous AI Research Hallucinates 90% Without Checks

rohanpaul_ai · x · 2026-09-01

A new Google paper reveals the risks of autonomous AI research. Without reliability modules, severe result hallucinations appeared in 90% of Agent Laboratory papers and 46% of Co-Scientist papers. When Co-Scientist cross-referenced claims with execution logs, the hallucination rate dropped to 4% and complete data fabrication to 0%.

Original post →

More from Safety

Safety channel →