Quantifying Gender Bias in Large Language Models: When ChatGPT Becomes a Hiring Manager

Collection of recent AI/ML papers addressing societal and safety impacts of large language and multimodal models. Key findings include: LLMs exhibit gender-related hiring biases (more likely to hire female candidates but recommending lower pay); deployed agent forums (Moltbook) show dramatically reduced threading and near-absence of author repair (author return 1.2% vs. 40.9% on Reddit), hampering public correction; a governance gap between nominal and genuine human oversight risks path-dependent lock-in with a 10–15 year governance window; an audit of Gemini 2.5 Pro multimodal search found 3.