Can AI be a Teaching Partner? Evaluating ChatGPT, Gemini, and DeepSeek across Three Teaching Strategies

This collection of arXiv papers (Mar 31, 2026) covers multiple AI and socio-technical topics with clear security and policy relevance. Key items: a position paper proposing "harmful capability uplift" as a human-centered metric for measuring how frontier models increase users' ability to cause harm (methodological guidance for evaluations); analyses of benchmarking risks for multi-agent scientific AI systems (contamination, tool-use, replication challenges); a federated, sector-led architecture for national AI governance and incident management (case study: India); empirical studies of LLMs as