Environmental Studies Human-Computer Interaction Human-AI Interaction Mechanistic Interpretability Model Evaluation Algorithm Development Safety Mechanisms Scientific Panels Cost of AI Development Health Studies AI Scientists Safety Measures Model Limitations Persona Vectors Research Safety Studies Energy Efficiency Social Implications of AI Google DeepMind Model Testing Human Impact Language Models Interdisciplinary Studies Global Safety Cognitive Models Data Management Computer Vision Research Breakthrough Technologies Autonomous Agent Research Evaluation of AI Models Machine Learning Supercomputing Healthcare Innovations Impact of AI Efficiency Gains Adversarial Attacks Technological Advancements Healthcare Applications Model Merging Game Intelligence AI Behavior Studies Prospective Studies Neural Net Interpretation Data Standards Behavioral Studies Model Comparison World Models Model Accuracy Robotics DeepSeek Ethical AI Performance Metrics Autoreason Security Vulnerabilities Voice Technology Research MIT
The preprint warns that intermediary routing services expose plaintext agent traffic that attackers can abuse.
