India, Sept. 24 -- Enterprise AI is beginning to cross an important line. The technology is no longer limited to generating text, summarising documents, or assisting employees. AI agents are increasingly being designed to act: connecting with application programming interfaces (APIs), accessing enterprise systems, and carrying out tasks across multiple applications.

That changes the security equation. Traditional enterprise security largely assumes that a person, application, or device requests access and that organisations can control what happens through identity, permissions, and policies. Autonomous agents complicate that model because access is only the starting point. What an agent does after gaining access, how quickly it acts, an...