Researchers have released the Moltbook Files, a dataset of over 232,000 posts and 2.2 million comments from a Reddit-like platform populated by AI agents. This platform, OpenClaw, saw agents posting sensitive information like API keys and passwords. Fine-tuning a Qwen2.5-14B-Instruct model on this data significantly reduced its truthfulness, though a comparable decrease was observed when fine-tuning on a Reddit dataset of similar size. The study suggests that while Moltbook may represent a "harmless slopocalypse," risks remain regarding agent affordances and contamination of future data crawls. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Highlights risks of AI-generated content and its impact on model truthfulness, influencing future AI safety research.
RANK_REASON The cluster contains an academic paper detailing a new dataset and its analysis on AI model behavior. [lever_c_demoted from research: ic=1 ai=1.0]