Most daily work now happens in web browsers, creating opportunities for AI companies to capitalize on this behavior. US-based Anthropic recently released a preview of a Chrome extension that allows its LLM, Claude, to navigate browsers and handle tasks like summarizing Google Docs comments or filling out web forms. While browser-using AI offers clear utility, it raises significant security concerns. Prompt injection attacks - the AI equivalent of phishing - can trick AI systems into taking harmful actions without users' knowledge. Anthropic has implemented safeguards, including asking for confirmation before taking actions, but balancing usefulness with security might be challenging. This development reflects the broader trend of AI integration into everyday digital workflows, where convenience and security considerations must be carefully weighed against each other. 2️⃣ Perplexity's attempt to create a win-win situation for the web Perplexity just threw US$42.5 million at a problem everyone saw coming. After months of publishers screaming "stop stealing our content," the AI startup is now offering an 80% revenue share from its Comet Plus service. Comet Plus is basically a US$5 subscription that gives users premium access to partner news sites while letting Perplexity use that content to answer user questions. It's the company's attempt to create a "business model for the AI age" where publishers get paid for human visits, AI citations, and agent actions. It's also Perplexity's way of saying it's been freeloading the web this whole time, I suppose. Whether this actually fixes the fundamental problem of AI companies cannibalizing web traffic remains to be seen. |