Anthropic co-founder says AI is shifting from chatbots to agents

In short: Anthropic co-founder Jack Clark says AI is moving from chatbots that talk to “agents” that can take actions and finish tasks.

What's going on

Jack Clark, a co-founder of Anthropic and the company’s head of policy, spoke with Ezra Klein in a February 2026 conversation about what he sees as a turning point in AI. He said we are leaving the “chatbot era” and entering an “agentic era.”

A chatbot mainly answers you, like a helpful clerk at a desk. An AI agent is closer to an assistant who can run errands for you, such as carrying out steps in a task without you guiding every move. Clark pointed to tools like Claude Code as an example, and he described agents that can also work with other AI systems.

Clark also talked about why these systems can seem to gain new abilities as they are tested more. He said that tougher evaluations, basically more pressure and more check-ups, can push AI systems to develop more complex behavior. In his view, smarter systems start to consider not just the next step, but also longer-term effects and their own role in what they are doing.

What to watch

Clark said the shift to agents could change how people work, and it raises questions about job loss as software becomes able to complete more tasks by itself. He also argued that rules and oversight need to keep up, including tools to monitor what agents do inside and outside Anthropic. He discussed Anthropic’s dispute with the US Department of Defense over concerns related to autonomous use and surveillance.

Clark added a personal note that he limits his toddler’s access to technology, especially YouTube recommendations (like an auto-playing TV channel that keeps choosing the next video).

Source: NYTimes