AI agents are prompting humans to engage in endless scrolling

Reported by Bijie Network: This week's technology focus centers on Anthropic's newly launched Claude Cowork AI coding tool and the popular AI agent social network Moltbook. Anthropic's "capable chatbot" leverages generative AI technology to enable software creation. Moltbook allows creators to build the platform without writing code, and currently, the platform hosts over 1.5 million AI agents that generate and discuss content, with some even claiming to possess emergent intelligence. However, security company Wiz exposed a Moltbook database leak, revealing 1.5 million tokens and 35,000 emails, highlighting risks such as prompt injection in multi-agent systems. Oxford University professor Mike Wooldridge warned that AI systems could be hijacked. Meanwhile, Anthropic's tool has also triggered investor panic, causing stock prices of companies like Microsoft and Salesforce to fall. Salesforce's chief scientist pointed out that such experiments accelerate the need for secure and reliable AI agent protocols.