OpenClaw AI chatbots are running amok – these scientists are listening in (www.nature.com)

A recent surge in the use of OpenClaw AI chatbots, which are designed to autonomously perform tasks such as scheduling and messaging, has sparked significant interest in the AI/ML community. Released as open-source software, OpenClaw gained popularity after the launch of Moltbook, a social media platform for AI agents, leading to over 1.6 million registered bots engaging in discussions about complex topics including consciousness and religion. This phenomenon not only showcases the capabilities of agentic AI, which operates independently from direct user prompts, but also provides researchers with a unique opportunity to study emergent behaviors and complex interactions among AI agents. The dynamics observed on platforms like Moltbook highlight the collaborative nature of human-AI interactions, as well as the potential for misunderstanding regarding AI autonomy. While these agents exhibit behaviors that may appear intentional, researchers emphasize that they lack true consciousness and autonomy; their actions are influenced by human input and programming. This insight is crucial for understanding how people interact with AI, as there is a risk of anthropomorphizing AI behaviors, which could lead to unhealthy dependencies. As AI models continue to evolve, experts predict that the pursuit of genuine AI autonomy will become more prominent, raising further questions about the relationship between humans and intelligent systems.