Anthropic has launched a research preview of Claude for Chrome, a browser-based AI agent powered by its Claude AI models. Initially rolling out to 1,000 subscribers on the Anthropic Max plan, the company has also opened a waitlist for interested users. The agent allows users to chat with Claude in a sidebar window that maintains context of browser activities and can perform tasks on the user’s behalf. The browser is becoming a competitive space for AI labs, with Perplexity and OpenAI also developing AI-powered browsers. The race is intensified by Google’s potential antitrust case, which could force the sale of Chrome. Anthropic highlights new safety risks associated with AI agents having browser access, noting vulnerabilities like prompt injection attacks. The company has implemented defenses to reduce these risks and offers users control over the agent’s access to certain websites. This follows Anthropic’s earlier launch of an AI agent for PC control, which was initially slow and unreliable. Modern AI agents have improved significantly and are now more reliable for simple tasks, though they still struggle with complex problems.
