In search of more data, Anthropic is asking users to 'help improve Claude' by providing the AI with access to their chatbot ...

AI agents that can control and read data from an internet browser are also susceptible to obeying malicious text circulating in web content.